Академический Документы
Профессиональный Документы
Культура Документы
INSTITUTO DE GEOFSICA
POSGRADO EN CIENCIAS DE LA TIERRA
TESIS
QUE PARA OBTENER EL GRADO DE:
DOCTOR EN CIENCIAS
(SISMOLOGA)
PRESENTA
DIRECTOR
Octubre de 2004
NDICE
Resumen
Introduccin
17
II. El sismo de Coyuca del 8 de Octubre del 2001, (Mw=5.8): Una falla normal
sobre la brecha ssmica de Guerrero.
30
46
56
64
Conclusiones
87
RESUMEN
Introduccin
Desde el sur del estado de Jalisco hasta el golfo de Tehuantepec los procesos
tectnicos regionales estn controlados por el rgimen convergente entre las placas de
Cocos y Norteamrica. De acuerdo al catlogo de sismos importantes del siglo XX
(Singh et al., 1984; Kostoglodov y Pacheco, 1999), la brecha ssmica de Guerrero se
encuentra acotada al noroeste por el rea de ruptura del sismo del 14 de Marzo de 1979
(Mw=7.4) (sismo de Petatln) (Valds et al., 1982) y al sureste por los sismos del 11 y 19
de Mayo de 1962 (Mw=7.1 y Mw=7.0, respectivamente) (Ortiz et al., 2000). El ltimo
temblor registrado en la brecha ssmica ocurri en Diciembre de 1911 (Ms 7.8). Previo a
este sismo ocurrieron los temblores de Diciembre de 1899 (Ms 7.7), Marzo de 1908 (Ms
7.8) y Julio de 1909 (Ms 7.5). Esta secuencia podra indicar que la regin libera energa en
eventos de magnitudes entre 7.5 y 7.8. Sin embargo, la dimensin total de la brecha ssmica
es capaz de generar un sismo de magnitud Mw=8.2 (Singh y Mortera, 1991). Dado que el
perodo estimado de retorno es de entre 60 y 70 aos la ocurrencia de un sismo importante
podra haberse retardado ms de 30 aos. En consecuencia, la brecha ssmica de Guerrero
es una de las regiones, de la zona de subduccin mexicana, con mayor probabilidad de
generar un temblor importante en los prximos aos.
Uno de estos temblores recientes es el sismo de Copalillo del 21 de Julio del 2000,
cuyo fallamiento normal confirma que el principal mecanismo que gobierna el estado de
esfuerzos en la placa de Cocos subducida es la fuerza de gravedad sobre la propia placa
("slab-pull"). En el trabajo titulado "Estudio de la fuente y de la propagacin del sismo
del 21 de Julio de 2000, Copalillo, Mxico(Mw=5.9): Implicaciones de los sismos
intraplaca en el peligro ssmico de la Ciudad de Mxico." (captulo I) se discuten las
principales caractersticas de este temblor. El mecanismo obtenido suponiendo una fuente
puntual muestra claramente que la naturaleza del temblor es extensiva, sin embargo no hay
a priori ninguna consideracin evidente que permita discriminar entre los dos planos de
falla resultantes de la solucin de mecanismo focal para una fuente puntual. Con el objeto
de discriminar entre ambos planos de falla, se llev a cabo una inversin cinemtica de la
ruptura usando datos locales y regionales a travs de un esquema propuesto originalmente
por Cotton y Campillo (1994) y modificado para hacer una bsqueda global eficiente
utilizando un esquema de recristalizacin simulada (simulated annealing) (Iglesias et al.,
2001).
Por otro lado, la ubicacin geogrfica de este temblor muestra que el peligro
ssmico asociado a los temblores de fallamiento normal podra ser importante para algunas
ciudades del Altiplano Mexicano incluyendo la propia ciudad de Mxico. En este trabajo se
discuten algunas de las implicaciones que este tipo de temblores podra tener en la
estimacin del riesgo ssmico para la ciudad de Mxico. La existencia de temblores
intraplaca de rgimen extensional en territorio mexicano no es poco comn. Solo la dcada
pasada ocurrieron dos sismos de este tipo (15-Jun-99, Mw=7.0 y 30-Sep-99, Mw=7.5) los
cuales causaron daos significativos en las ciudades de Puebla (Singh et al., 1999) y
Oaxaca, respectivamente (Singh et al., 2000, Hernndez et al., 2001).
Otro evento de especial inters ocurri el 8 de Octubre del 2001 (Mw=5.8) muy
cerca de la poblacin de Coyuca de Bentez, Gro. Este sismo tambin fue asociado a un
mecanismo de falla normal. Sin embargo, la profundidad hipocentral indica claramente que
no ocurri dentro de la placa de Cocos subducida, sino en la placa cabalgante de
Norteamrica. El epicentro de este sismo se encuentra localizado precisamente sobre la
"brecha ssmica" de Guerrero. Dada su posicin con respecto a la zona acoplada (justo
encima), este temblor es, probablemente, el nico evento de esta naturaleza y magnitud
reportado hasta ahora.
En el trabajo "El sismo de Coyuca del 8 de Octubre del 2001, (Mw=5.8): Una
falla normal sobre la brecha ssmica de Guerrero" (captulo II), se analizan las
caractersticas principales del sismo de Coyuca y sus implicaciones en la tectnica regional.
Numerosas rplicas fueron localizadas con bastante precisin gracias a una red local de
estaciones porttiles desplegadas despus del sismo principal. Las rplicas muestran
claramente la orientacin del plano de falla. Una inversin cinemtica de la ruptura, usando
el esquema anteriormente mencionado, permiti conocer a groso modo las caractersticas
generales de la ruptura.
S bien algunos autores (P.ej. Singh y Pardo, 1993) han propuesto que la placa
cabalgante de Norteamrica se encuentra en un rgimen tensional, hasta ahora la evidencia
provena de sismos pequeos continentales lejanos a la costa. Este rgimen tensional,
evidenciado por el sismo de Coyuca, podra estar relacionado al fenmeno llamado
"retroceso de la trinchera" (P.ej. Molnar y Atwater, 1978; Uyeda y Kanamori, 1979;
Nakamura y Uyeda, 1980) o al conocido como "erosin tectnica" (P.ej. Murauchi, 1971).
Poco despus del sismo de Coyuca del 8 de Octubre, comenz en la misma regin
un deslizamiento extremadamente lento y assmico (Kostoglodov et al., 2003). Este
deslizamiento fue registrado solamente por las estaciones permanentes GPS del
Departamento de Sismologa del Instituto de Geofsica de la UNAM. El deslizamiento
assmico mencionado ocurri desde principios del mes de Enero del 2002 y tuvo una
duracin de al menos 4 meses. El desplazamiento tuvo lugar muy probablemente en la
interfase de las placas de Cocos y Norteamrica.
Trabajos previos consideran que el lmite superior del ancho de la zona acoplada,
para la zona de subduccin mexicana, es menor a ~80 Km. (P.ej Singh et al.,1985, Singh y
Mortera, 1991). En contraste, el modelo propuesto en el presente trabajo considera una
zona acoplada ms ancha (~150Km). Debido a que parte de esta zona acoplada libera
10
Los resultados de las inversiones muestran, tambin, que la parte del modelo ms
cercana a la trinchera tiene un bajo grado de acoplamiento lo cual no significa que no
existan sismos de pequea o moderada magnitud en esta zona. Precisamente, el 18 de Abril
del 2002 ocurri un sismo de Mw=6.7 a unos 55 km frente a las costas del estado de
Guerrero. Esta distancia lo ubica en una zona muy cercana a la trinchera mesoamericana.
Por esta razn, este sismo es catalogado como un sismo de trinchera. Este tipo de sismos
merece un tratamiento especial ya que presentan caractersticas peculiares tanto en las
cualidades de la fuente ssmica como en las trayectorias de propagacin.
Shapiro et al. (1998) analizaron este tipo de sismos en Mxico, encontrando que su
espectro de Fourier presenta consistentemente deficiencia de amplitud en altas frecuencias
con respecto a los sismos que ocurren cercanos a la costa. Diversos estudios (P.ej. Ide et
al., 1993; Kanamori y Kikcuhi, 1993) han resaltado la relacin entre este tipo de sismos y
la ocurrencia del fenmeno conocido como Tsunami. Con base en lo anterior, Shapiro et al.
(1998) propusieron la implementacin de un sistema rpido de alerta de Tsunami basada en
la discriminacin entre un sismo de trinchera y un sismo de otra naturaleza a travs de un
simple anlisis del espectro del temblor. En el artculo titulado Los sismos de trinchera
en Mxico presentan aceleraciones mximas anmalamente bajas (captulo IV), se
analizan las caractersticas del sismo del 18 de Abril catalogndolo como un sismo
potencialmente tsunamignico (utilizando el criterio antes mencionado). En este trabajo se
retoman las observaciones hechas por Shapiro et al. (1998) acerca de la relacin entre la
deficiencia de amplitud en altas frecuencias y este tipo de sismos, para mostrar que las
aceleraciones mximas que provocan los sismos de trinchera son consistentemente menores
a las aceleraciones mximas provocadas por sismos "costeros". Esta conclusin, junto con
la presentada por Shapiro et al. (1998), conducen al planteamiento de una paradoja; por un
lado este tipo de sismos tsunamignicos representan un peligro para las poblaciones
costeras, pero por otro lado, las aceleraciones que se producen son menores a las
11
provocadas por sismos costeros y por tanto los daos esperados en las estructuras pueden
ser menores.
Dado que las ondas ssmicas viajan a mucha menor velocidad que las
electromagnticas, la seal emitida en las estaciones llega aproximadamente 50 segundos
antes que las ondas ssmicas de gran amplitud. De acuerdo al "Centro de Instrumentacin y
Registro Ssmico" (CIRES) (Espinosa-Aranda et al., 1997), el sistema de alerta dispara en
modo restringido (solamente para algunas instituciones) con sismos de magnitud 5M<6 y
en alerta pblica (estaciones de radio) para sismos de M 6. Sin embargo, en el caso del 18
de Abril, a pesar de que el sismo tuvo una magnitud M 6, el SAS no dispar ni siquiera
en alerta restringida. El CIRES report que no se detect ninguna falla en el sistema y que
la razn por lo cual no se activ la alerta se encontraba en que en ninguna de las estaciones
que registro el evento, haba superado el umbral para tal propsito. Esto concuerda con las
conclusiones alcanzadas en el ya mencionado captulo IV de esta tesis.
12
anlisis del desempeo del SAS. Este anlisis fue hecho en funcin de los sismos que
generaron tanto alerta pblica como restringida y las aceleraciones mximas que
provocaron en la Ciudad de Mxico. En este sentido el SAS muestra un desempeo
deficiente bsicamente por dos razones:
Debido a que el SAS ha sido diseado para disparar solamente con los sismos
localizados a lo largo de la costa de Guerrero, el sistema no es capaz de detectar sismos que
provienen de otras zonas con alto potencial ssmico y que tambin pueden afectar a la
Ciudad de Mxico. Por esta razn es deseable evaluar la posibilidad de extender el sistema
para abarcar la mayor parte de las regiones cuyo potencial sismognico represente un riesgo
para la Ciudad de Mxico.
Siguiendo el esquema del actual sistema, esta tarea pudiera verse complicada tanto
desde el punto de vista logstico como econmico, ya que se requeriran un gran nmero de
estaciones y, por lo tanto, un gran esfuerzo en las fases de instalacin, operacin y
mantenimiento.
Una de las partes crticas de una alerta temprana es el mtodo para determinar en
muy poco tiempo s el sismo puede causar daos o no. En este contexto diversos autores
han trabajado en el diseo de algoritmos que en solo unos cuantos segundos pueden
discriminar entre los eventos pequeos y los grandes (P. ej . Tsuboi y Kikuchi, 2002; Allen
y Kanamori,2003).
13
Los resultados obtenidos muestran que es posible separar, de manera confiable, los
eventos que exceden 2 gales en la estacin CUIP usando el valor de RMS filtrado en la
estacin cercana. Este umbral parece adecuado ya que, al menos, la mayora de las personas
sentiran cualquier evento superando ese umbral.
Por otro lado esta propuesta presenta un mejor desempeo, con respecto a falsas
alertas y fallos (eventos importantes para los cuales no dispar el sistema).
14
Referencias:
Allen R. y H. Kanamori, 2003. The potential for earthquake early warning in Southern California. Science. 300,
786-789.
Cotton, F. y M. Campillo,1995. Inversion of strong ground motion in the frequency domain, J. Geophys. Res.
100, 3961-3975.
Converse, G., 1973. Equations for the displacements and displacement derivatives due to a rectangular
dislocation in a three-dimensional elastic half-space, unpublished, U. S. Geological Survey, Menlo Park.
Erickson, L., 1986. A three-dimensional dislocation program with applications to faulting in the Earth, M.S.
thesis, 167 pp., Stanford Univ., Stanford, Calif.
Espinosa Aranda, J.M:, A. Jimnez, G. Ibarrola, F. Alcantar, A. Aguilar, M. Inostroza y S. Maldonado,1995.
Mexico City seismic alert system, Seism. Res. Lett. 66, 42-53.
Fedotov, S.A., 1965. Regularities of the distribution of strong earthquakes in Kamchatka, the Kurile islnads and
northeastern Japan, Trans. Acad. Sci. USSR, Inst. Phys. Earth 36, 66-93.
Hernandez, B., N. Shapiro, S.K. Singh, J. Pacheco, F. Cotton, M. Campillo, A. Iglesias, V. Cruz, J.M. Gmez y
L. Alcntara, 2001. Rupture history of Sep 30, 1999 intraplate earthquake of Oaxaca, Mexico (Mw=7.5) from
inversion of strong motion data in the frequency domain, Geophys. Res. Lett., 28, 363-366.
Iglesias, A., V.M. Cruz-Atienza, N.M. Shapiro, y J.F. Pacheco, 2001. Crustal structure of south-central Mexico
estimated from the inversion of surface-wave dispersion curves using genetic and simulated annealing
algorithms, Geof. Int., 40, 181-190.
Ide, S, F. Imamura, Y. Yoshida y K. Abe, 1993. Source characteristics of the Nicaraguan tsunami earthquake
of September 2, 1992, Geophys. Res. Lett. 9, 863-866.
Kanamori, H. y M. Kikuchi, 1993. The 1992 Nicaragua earthquake: a slow tsunami earthquake associated with
subducted sediments, Nature 361, 714-716.
Kanamori, H., P.C. Jennings, S.K. Singh y L. Astiz, 1993. Estimation of strong ground motion in Mexico City
expected for large earthquakes in the Guerrero seismic gap, Bull Seism. Soc. Am., 83, 811-829.
Kelleher, J., Sykes, L. y Oliver, J., 1973. Possible criteria for predicting earthquake locations and their
applications to major plate boundaries of the Pacific and the Caribbean. J. Geophys. Res., 78, 2547-2585.
Kostoglodov, V., Larson K. M., S. K. Singh, A. Lowry, J. A. Santiago, S. I. Franco y R. Bilham, 2003. A large
silent earthquake in the Guerrero seismic gap, Mexico. Geophys. Res. Lett. 30, 1807.
Kostoglodov V., Pacheco J. Un catlogo de sismos moderados y grandes ocurridos en Mxico durante el siglo
XX, Poster 100 aos de sismicidad en Mxico, Instituto de Geofsica, UNAM, 1999.
Lowri, A. R., Larson K. M. Kostoglodov, V. and Bilham,R., 2001. Tansient fault slip in Guerrero, southern
Mexico. Geophys, Res. Lett. 28, 3753-3756 .
McCann, W. R., S. P. Nishenko, L. R. Sykes y J. Krause, 1979. Seismic gaps and plate tectonics: Seismic
potential for major boundaries, Pure Appl. Geophys. 217, 1082-1147.
Molnar, P. y T. Atwater, 1978. Interarc spreading and cordilleran tectonics as alternates related to the age of
subducted oceanic lithosphere, Earth Planet, Sci. Lett.. 330-340, 1978
Murauchi, S., 1971. The renewal of island arcs and the tectonics of marginal seas, 1971. En Island Arcs and
Marginal Seas. Tokai University Press, Tokyo.
Nakamura K. y S. Uyeda, 1980. Stress gradient in arc-back arc regions and plate subduction. J. Geophys.
Res., 85, 6419-6428.
15
Nishenko, S.P. y S.K. Singh, 1987, Conditional probabilities for the recurrence of large and great interplate
earthquakes along the mexican subduction zone. Bull. Seism. Soc. Am., 77, 2095-2114.
Okada,Y., 1992. Internal deformation due to shear and tensile faults in a half-space,.
Am., 82, 1018-1040.
Bull.
Seism.
Soc.
Ordaz, M. J. Arboleda y S.K. Singh, 1995. A scheme of random summation of an empirical Green's function to
estimate ground motions from future large earthquakes, Bull. Seism. Soc. Am., 85, 1635-1647.
Ortiz, M., S. K. Singh, V. Kostoglodov y J. Pacheco, 2000. Source areas of the Acapulco-San Marcos, Mexico
earthquakes of 1962 (M 7.1; 7.0) and 1957 (M 7.7), as constrained by tsunami and uplift records. Geof. Int.,
39, 337-348.
Shapiro, N.M., S.K. Singh y J. Pacheco, 1998. A fast and simple diagnostic method for identifying
tsunamigenic earthquakes, Geophys. Res. Lett. 25, 3911-3914.
Singh, S. K., L. Astiz y J. Havskov, 1981. Seismic gaps and recurrence periods of large earthquakes along the
Mexican subduction zone: a Reexamination. Bull. Seism. Soc. Am., 71, 827-843.
Singh, S.K., M. Rodrguez y J. M. Espndola, 1984. A catalog of earthquakes of Mexico from 1900 to 1981,
Bull. Seism. Soc. Am. 74, 267-279.
Singh, S.K., G. Surez y T. Domnguez, 1985. The great Oaxaca earthquake of 15 January 1931: Lithosphere
normal faulting in the subducted Cocos plate. Nature, 317, 56-58.
Singh, S.K. y F. Mortera, 1991. Source-time functions of large Mexican subduction earthquakes, morphology of
the Benioff zone and the extent of the Guerrero gap, J. Geophys. Res., 96, 21487-21502.
Singh, S.K. y M. Pardo, 1993. Geometry of the Benioff Zone and state of stress in the overriding plate in
Central Mexico. Geophys. Res. Lett., 20, 1483-1486.
Singh, S.K., M. Ordaz, J.F. Pacheco, R. Quaas, L. Alcntara, S. Alcocer, C. Gutirrez, R. Meli y E.
Ovando,1999. A preliminary report on the Tehuacn, Mxico earthquake of June 15, 1999 (Mw=7.0), Seism.
Res. Lett. 70, 489-504.
Singh, S.K., M. Ordaz, L. Alcntara, N. Shapiro, V. Kostoglodov, J. F. Pacheco, S. Alcocer, C. Gutierrez, R.
Quaas, T. Mikumo y E. Ovando, 2000. The Oaxaca Earthquake of September 30, 1999 (Mw=7.5): A NormalFaulting Event in the Subducted Cocos Plate, Seism. Res. Lett. 71, 67-78.
Tsuboi, S.; Saito y Kikuchi, M., 2002. Real-time earthquake warning by using broadband P Waveform.
Geophys. Res. Lett., 29, 1483-1486
Uyeda, S. y H. Kanamori, 1979. Back arc opening and the mode of subduction, J. Gepohys. Res., 84, 10491061.
Valds, C., R.P. Meyer, R. Zuiga, J. Havskov and S.K. Singh, 1982. Analysis of Petatlan aftershocks:
Numbers energy release, and asperities, J. Geophys. Res., 87, 8519-8527.
16
Captulo I
17
Bulletin of the Seismological Society of America, Vol. 92, No. 3, pp. 10601071, April 2002
Abstract
The Copalillo earthquake of 21 July 2000 (Mw 5.9) is the closest, welllocated inslab event to Mexico City ever to be recorded. In this study, we analyze
local and regional broadband and accelerometric recordings to determine the source
parameters of the earthquake and the attenuation of ground motion with distance and
to obtain a preliminary estimate of the seismic hazard posed to the city by such
events. Our results show that the earthquake occurred at a depth of about 50 km,
most probably in the subducted oceanic crust. The waveform inversion discriminates
between the two nodal planes; the fault plane defined by the following: strike, 305;
dip, 32; and rake, 80. The rupture propagated nearly unilaterally along the strike
toward northwest with a small downdip component. The observed source spectrum
can be well explained by an x2-source model with M0 6.0 1025 dyne cm and
a stress drop of 360 bar. We find that high-frequency ground motion (f 3 Hz),
which is related to Amax during inslab earthquakes, is not amplified at Ciudad Universitaria (CU), a hill-zone site in the Valley of Mexico that is known to suffer amplification at low frequencies (0.1 f 2.0 Hz). Simulations using the recording
at CU of the Copalillo earthquake as an empirical Greens function suggests that a
Mw 7.0 event could give rise to an Amax value of 3040 gal. The CU recordings
indicate that the Amax value of 30 gal could have a return period of about 40 yr, about
the same as from shallow-dipping thrust earthquakes along the Mexican subduction
zone, which have been regarded as posing the highest hazard for the city. An inslab
earthquake with an Amax value of about 40 gal could cause heavy damage to small
buildings at certain locations of the city. We conclude that seismic hazard from inslab
earthquakes to Mexico City has so far been underestimated.
Introduction
Inslab earthquakes in the subducted Cocos plate below
south-central Mexico cease to occur at depths of less than
about 80 km and well before reaching the Mexican Volcanic
Belt (MVB). Indeed, no well-located inslab earthquake is
known to have occurred below the MVB. The recent earthquake of 21 July 2000 (Mw 5.9; H 50 km), henceforth
called the Copalillo earthquake, is the closest, reliably located, normal-faulting inslab event to Mexico City (Fig. 1).
The epicenter of the earthquake lies near the town of Copalillo, about 65 km to the northeast of Iguala and 136 km
to the south of Ciudad Universitaria (CU), Mexico City. In
this article, we study the characteristics of the source and
the ground motions generated by this earthquake and discuss
the implications for seismic hazard to the city from inslab
earthquakes. There are several reasons for this endeavor.
(1) Source characteristics of inslab earthquakes in the Cocos
plate, while essential to understand the mechanism of generation of such events, are available for only a few earthquakes. (2) There is evidence that high-frequency ground
motion from inslab events in Mexico are more intense than
from interplate earthquakes (see, e.g., Singh et al., 2000).
This observation needs further confirmation. (3) A critical
issue in the estimation of seismic hazard to Mexico City is
related to inslab, normal-faulting earthquakes in the subducted Cocos plate.
The concern about the seismic hazard arises because in
the past inslab earthquakes have caused significant damage
to cities and towns in the Mexican altiplano. There are many
recent examples. The earthquake of 15 January 1931 (M 7.8,
H 40 km) caused severe destruction to the City of Oaxaca
(Barrera, 1931; Singh et al., 1985); the earthquakes of 28
August 1973 (Mw 7.0; H 82 km) and 24 October 1980
(Mw 7.0; H 65 km) resulted in deaths and damage in the
states of Veracruz, Puebla, and Oaxaca (Singh and Wyss,
1060
18
A Source and Wave Propagation Study of the Copalillo, Mexico, Earthquake of 21 July 2000 (Mw 5.9)
1061
Figure 1. (Top) Epicenters of large (M 6.9) inslab, normal-faulting, intermediatedepth earthquakes and location of Mexico City. Also shown is the 21 July, 2000 Copalillo earthquake (Mw 5.9) (star). Lightly shaded area represents the Mexican Volcanic
Belt (MVB). (b) The projection of the hypocenters of the inslab events on a vertical
plane AA. The dashed lines delineate the subducted Cocos plate. The depths of the
nineteenth-century earthquakes (dashed circles) have been arbitrarily assigned to fall
near to the plate interface. The 1937 event lies above the plate interface probably
because of the error in its depth.
1976; Lomnitz, 1982; Yamamoto et al., 1984; Nava et al.,
1985). The recent earthquake of 15 June 1999 (Mw 6.9, H
60 km) caused damage in the State of Puebla, especially to
colonial structures in and near the City of Puebla (Singh et
al., 1999). The 30 September 1999 Oaxaca earthquake (Mw
7.4; H 40 km) was damaging to the City of Oaxaca and
many towns along the coast and between the coast and the
city of Oaxaca (Singh et al., 2000). It has been suggested
19
1062
to Mexico City and its likely maximum magnitude are uncertain. The design spectra for the Federal District does contemplate an inslab Mw 6.5 earthquake at focal distance of 80
km from Mexico City (Rosenblueth et al., 1989). The choice
of the magnitude and the focal distance were based on the
available data at that time and the best judgment of the authors. The expected ground motion from such an earthquake
was estimated assuming an x2 seismic source model and
application of results from random vibration theory (Boore,
1983), with little constraint from actual recordings. For these
reasons, the seismic hazard to Mexico City from inslab
earthquakes is subject to relatively large uncertainty.
Figure 1 shows epicenters of large (M 6.9), inslab
earthquakes in the subducted Cocos plate. It should be noted
that the locations and the magnitudes of nineteenth-century
events are less reliable. Figure 1 suggests that such earthquakes can reach a magnitude of 7.3 within 200 km of Mexico City. The figure also includes the location of the Copalillo earthquake. As mentioned earlier, this earthquake is the
closest inslab earthquake. In view of Figure 1, an inslab Mw
7.3 earthquake may reasonably be expected to occur as close
as 136 km from the city at a depth of about 50 km. It is,
therefore, of interest to estimate ground motions in the Valley of Mexico from such an earthquake. We do this by using
a CU recording of the Copalillo earthquake as the empirical
Greens function (EGF) to simulate expected ground motion
in the city from future large earthquakes in the same region.
We show that the point-source approximation inherent in the
EGF method is quite reasonable and the effect of source
directivity on the estimation of Amax in CU is not significant
(less than 8%) for postulated earthquakes with Mw 7.5. We
finally discuss the significance of the results in terms of seismic hazard to Mexico City from inslab earthquakes.
stituto de Geofisica (IGF) and from the stations of the accelerographic networks operated by Instituto de Ingeniera
(II) and Centro Nacional de Prevencion de Desastres (CENAPRED). A typical BB station of SSN consists of a STS-2
seismometer and a Kinemetrics FBA-23 accelerometer connected to a 24-bit Quanterra digitizer. Continuous velocity
data, sampled at 1 and 20 Hz, are saved in a buffer memory.
For triggered events both velocity and acceleration channels,
sampled at 80 Hz, are saved. The accelerometric networks
mostly consist of Kinemetrics K2 and ETNA digital accelerographs, equipped with 19 and 18 bit A to D converters,
respectively. The time synchronization is provided by GPS
receivers. The triggered events are saved at a sampling rate
of 200 Hz. Descriptions of the accelerometric networks may
be found in Anderson et al. (1994) and Quass et al. (1987,
1989, 1993).
Source Parameters
The location of the event, determined from local and
regional data, and the source parameters, obtained from the
moment tensor (MT) inversion of regional waveforms, are
given in Table 2. The MT solution was obtained from the
inversion of the bandpass-filtered (between 20 and 50 sec)
BB regional seismograms. The procedure of Randall et al.
(1995) was followed in the inversion. The details of the
method, as applied to the analysis of BB data of the Mexican
network, are given by Pacheco and Singh (1998). The crustal
structure used in the location and the waveform modeling is
given in Table 3. The source parameters reported in the Harvard centroid moment tensor (CMT) catalog are included in
Table 2
Source Parameters of the 21 July 2000, Copalillo Earthquake
Data
The data used in the analysis of the Copalillo earthquake
consist of recordings from the broadband (BB) seismic network of the Servicio Sismologico Nacional (SSN) of the In-
Table 1
Significant Inslab Earthquakes (M 6.9) in the Subducted Cocos
Plate close to Mexico City
Event
Date
(yyyymmdd)
Latitude
Longitude
Depth
Magnitude
1
2
3
4
5
6
7
8
9
10
11
18580618
18641003
18790517
19310115
19370726
19640706
19730828
19801024
19970111
19990615
19990930
18.00
18.70
18.60
16.34
18.48
18.31
18.00
17.90
18.06
18.15
16.00
100.80
97.40
98.00
96.87
96.08
100.50
96.55
98.15
102.79
97.52
97.02
40
85
55
82
65
34
60
40
7.7
7.3
7.0
7.8
7.3
7.3
7.0
7.0
7.1
6.9
7.4
Source
Regional*
Regional
CMT
Latitude
(N)
Longitude
(E)
18.113
18.28
98.974
98.77
Depth
(km)
M0
dyne cm
50.0
58.4
4.1 10
6.0 1024
7.9 1024
24
Strike
Dip
Rake
305
310
32
39
80
75
Table 3
Crustal Structure Used in Locating and Modeling
the Copalillo Earthquake
Layer Thickness
(km)
P-Wave Speed
(km/sec)
S-Wave speed
(km/sec)
Density
(gm/cm3)
7.7
12.0
23.3
36.7
5.60
6.00
6.90
8.10
8.40
3.20
3.41
3.92
4.67
4.85
2.56
2.69
2.98
3.36
3.46
20
A Source and Wave Propagation Study of the Copalillo, Mexico, Earthquake of 21 July 2000 (Mw 5.9)
1063
Figure 2.
21
1064
Figure 3.
22
A Source and Wave Propagation Study of the Copalillo, Mexico, Earthquake of 21 July 2000 (Mw 5.9)
0 ( f ) epfRi /b Q( f )/G(Ri),
Ai ( f, Ri) C f 2 M
(1)
(2)
1065
where
(3)
(4)
(5)
Figure 5.
23
1066
100
Ob s e r v e
d A m ax at CU ( g al )
ed
10
0.1
0.1
10
100
Pr e d i c t e d A m ax ( g al )
Table 4
Peak Accelerations Recorded at CU during Inslab Earthquakes in the Subducted Cocos Plate
Date
(yyyymmdd)
Latitude
(N)
Longitude
(E)
Depth
(km)
Mw
R
(km)
Amax (N)
(gal)
Amax (E)
(gal)
Amax (Z)
(gal)
19640706
19730828
19801024
19930805
19940223
19940506
19940523
19941210
19970111
19970403
19970522
19980420
19990615
19990621
19990930
19991229
20000721
18.310
18.250
18.030
17.429
17.750
18.390
18.020
17.982
18.340
18.510
18.370
18.350
18.130
18.150
16.030
18.000
18.120
100.500
96.550
98.270
98.337
97.270
97.980
100.570
101.517
102.580
98.100
101.820
101.190
97.540
101.700
96.960
101.630
98.970
55
82
65
54
75
57
50
49
40
52
54
64
60
53
40
50
49
7.3
7.0
7.0
5.2
5.8
5.2
6.2
6.4
7.1
5.2
6.5
5.9
6.9
6.3
7.4
5.9
5.9
179
301
173
229
267
163
206
288
374
145
298
238
218
296
435
297
136
18.30
(19.52)*
25.30
0.54
1.04
0.44
5.00
5.40
4.20
1.20
1.70
1.30
11.90
3.10
7.80
1.19
12.21
15.70
(17.54)*
23.50
0.46
0.95
0.54
4.30
5.50
5.90
2.00
2.10
1.50
11.40
3.40
7.70
1.38
9.35
12.00
(12.87)*
12.50
0.35
0.62
0.40
2.90
2.60
3.10
1.20
1.70
1.20
7.50
1.60
5.10
0.73
5.80
*The earthquake was not recorded at CU, so it was not used in the analysis. The peak values listed here
correspond to station Palacio de los Deportes.
24
A Source and Wave Propagation Study of the Copalillo, Mexico, Earthquake of 21 July 2000 (Mw 5.9)
Figure 7.
1067
time histories whose envelopes are realistic. The method requires specification of the seismic moments and the stress
drops of the EGF and the target event. The details of the
method are given in Ordaz et al. (1995). If only peak ground
motion parameters are desired, then the computation of the
time histories is bypassed; the Fourier spectrum, along with
an estimation of duration (TR) of the intense part of the
ground motion, and application of results from random vibration theory (RVT) suffices (see Appendix B of Ordaz et
al., [1995] for relevant formulas).
In the synthesis, we take Dr 360 bar for both the
EGF and the target event. The duration TR in sec is given
by TR f c1 0.05R, where f c is the corner frequency
(equation 4) and R is the hypocentral distance in km (Hermann, 1985). The results of the synthesis are shown in Figure 8. We note that for a Mw 7.0 earthquake in the Copalillo
region (and probably also in any other region at similar focal
distance from CU, except in sector 1), the expected peak
acceleration (Amax), velocity (Vmax), and displacement (Dmax)
at CU are about 35 gal, 7 cm/sec, and 4 cm, respectively.
The corresponding values for a Mw 7.3 earthquake are about
50 gals, 10 cm/sec, and 7 cm, respectively. The predicted
Amax as function of Mw based on regression analysis (equation 5) is also shown in Figure 8 (straight line). The predicted
Amax from the EGF technique and the regression analysis are
in very good agreement. Figure 9 illustrates samples of time
histories (northsouth component). Note that the waveforms
are quite realistic.
We recall that the previous synthesis is based on a pointsource approximation. For this approximation to be valid, L
and k should be smaller than R, where L is the source dimension, R is the hypocentral distance, and k is the wavelength of interest. In our case, R 145 km, and k 20 km.
Let us assume that the rupture area, A, is square in shape. A,
in km2, is related to Mw by the following relation: log A
Mw 4.0 (e.g., Wyss, 1979; Singh et al., 1980). It follows
that L is about 32 km, 45 km, and 56 km for Mw 7.0, 7.3,
and 7.5 earthquakes, respectively. These estimations suggest
that for Mw 7.5 events, which are of interest in this study,
the point-source approximation may be valid. We further
tested the validity of this approximation by using finitesource stochastic model of Beresnev and Atkinson (1997,
1998, 1999, 2001). The method requires the fault plane to
be divided in subfaults whose size, Dl, in km, is given by
log Dl 0.4 Mw 2.0. The subfaults are stochastic x2
sources. The subevent time history at a site is generated following the procedure of Boore (1983). The rupture propagates radially from a specified hypocenter, and the randomness is introduced in the subevent rupture times. A standard
technique sums the contribution from each subfault. A stress
parameter, which relates subfault moment and its size, is
fixed at 50 bars. A free parameter, called the strength factor,
sfact, which controls the level of high-frequency radiation,
needs to be specified (see Beresnev and Atkinson, 1997,
1998). Other required parameters are the same as used in the
estimation of the source spectrum. The azimuth of the fault
25
1068
Figure 8. Expected peak ground motions at CU from postulated future large earthquakes in the region of Copalillo. The Copalillo earthquake recording has been used
as empirical Greens function (EGF). A stress drop of 360 bars has been taken for both
the EGF and the target event. The straight line in the left frame is the predicted Amax
value based on regression relation given in equation (5).
Figure 9. Samples of simulated peak ground-motions at CU (northsouth component) at CU from Mw 6.5 and 7.0 earthquakes. The Copalillo recordings, used as EGF,
are also shown.
and its dip were taken as 305 and 32, respectively (Table
2). We fixed the depth of the upper edge of the fault to 50
km. We first estimated the radiation strength parameter,
sfact, to be 2 by requiring that the predicted values of Amax
for Mw 7.0 from point-source EGF method and finite-source
stochastic method (with hypocenter located in the center of
the rupture area) be the same. This value of sfact is reasonable in view of the large Brunes stress drop, 360 bars, for
the Copalillo event. We estimated Amax using the finitesource model for postulate earthquakes of Mw 7.3 and 7.5
and found these values were almost identical with those predicted by the point-source EGF method. This further lends
weight to the validity of the point-source approximation.
Finally, we estimated the effect of the source directivity on
Amax at CU. This effect is not included in the results of
EGF simulation. To accomplish this goal, we again used the
finite-source stochastic model and simulated ground motions
26
A Source and Wave Propagation Study of the Copalillo, Mexico, Earthquake of 21 July 2000 (Mw 5.9)
1069
Figure 10.
27
1070
Conclusions
References
Acknowledgments
We thank Igor Beresnev for providing us with his finite-source stochastic simulation program. The research was partially funded by DGAPA,
UNAM Project IN109598 and CONACyT Projects 25403-A and 26185-T.
28
A Source and Wave Propagation Study of the Copalillo, Mexico, Earthquake of 21 July 2000 (Mw 5.9)
and from MT inversion of regional data: a comparison of results, Geofis. Intern. 37, 95102.
Quaas, R., J. G. Anderson, and D. Almora (1987). La red acelerografica de
Guerrero para registro de temblores fuertes, in Memoria VII Congreso
Nacional de Ingeniera Ssmica, Queretaro, Mexico, 1921 November, B40B54.
Quaas, R., J. G. Anderson, and D. Almora (1989). La red acelerografica de
Guerrero, 4 anos de operacion, Ingeniera Ssmica 36, 5368.
Quaas, R., J. A. Otero, S. Medina, J. M. Espinosa, H. Aguilar, and M.
Gonzalez (1993). Base Nacional de Datos de Sismos Fuertes, Catalogo de Estaciones Acelerograficas 19601992, Soc. Mex. Ingeniera
Ssmica, A. C., Mexico, 310 pp.
Randall, G. E., C. J. Ammon, and T. J. Owens (1995). Moment tensor
estimation using regional seismograms from a Tibetan plateau portable network deployment, Geophys. Res. Lett. 22, 16651668.
Reinoso, E., and M. Ordaz (1999). Spectral ratios for Mexico City from
free-field recordings, Earthquake Spectra 15, 273296.
Rosenblueth, E., M. Ordaz, F. J. Sanchez-Sesma, and S. K. Singh (1989).
Design spectra for Mexicos Federal District, Earthquake Spectra 5,
273292.
Shapiro, N., S. K. Singh, A. Iglesias-Mendoza, V. Cruz-Atienza, and J. F.
Pacheco (2000). Evidence of low Q below Popocatepetl, and its implication to seismic hazard in Mexico City, Geophys. Res. Lett. 27,
27532756.
Singh, S. K., and M. Wyss (1976). Source parameters of the Orizaba earthquake of August 28, 1973, Geofis. Internacional 16, 165184.
Singh S. K., E. Bazan, and L. Esteva (1980). Expected earthquake magnitude from a fault, Bull. Seism. Soc. Am. 70, 903914.
Singh, S. K., J. Lermo, T. Domnguez, M. Ordaz, J. M. Espinosa, E. Mena,
and R. Quaas (1988a). A study of relative amplification of seismic
waves in the valley of Mexico with respect to a hill zone site (CU),
Earthquake Spectra 4, 653674.
Singh, S. K., E. Mena, and R. Castro (1988b). Some aspects of source
characteristics of 19 September 1985 Michoacan earthquake and
ground motion amplification in and near Mexico city from the strong
motion data, Bull. Seism. Soc. 78, 451477.
Singh, S. K., M. Ordaz, L. Alcantara, N. Shapiro, V. Kostoglodov, J. F.
Pacheco, S. Alcocer, C. Gutierrez, R. Quaas, T. Mikumo, and E.
1071
29
Captulo II
30
El sismo de Coyuca del 8 de Octubre del 2001, (Mw=5.8): Una falla normal
sobre la brecha ssmica de Guerrero.
A. Iglesias, J.F. Pacheco y S. K. Singh
Instituto de Geofsica, Universidad Nacional Autnoma de Mxico, Mxico, D. F. Mxico
Introduccin
El 8 de Octubre del 2001, un sismo de magnitud moderada (Mw=5.8) fue registrado
cerca de la poblacin de Coyuca de Bentez en el estado de Guerrero. Esta poblacin est
localizada cerca de la costa y justo encima de la brecha de Guerrero (figura 1). La
profundidad reportada para este sismo es de ~ 8 Km. La interfase entre las placas de
Norteamrica (NOAM) y la placa de Cocos subducida se encuentra a unos 20 Km de
profundidad, por lo que el sismo de Coyuca se sita en la placa cabalgante de Norteamrica
(figura 2).
El ltimo temblor de subduccin importante en la zona ocurri en 1911 (16 de
diciembre de 1911, M=7.5) y, dado que el periodo de recurrencia de la regin ha sido
estimado entre 60 y 70 aos para sismos de magnitud 7.7 (Nishenko y Singh 1987), la
ocurrencia de un temblor en la regin podra haberse retardado por ms de 30 aos. Por esta
razn esta zona es considerada como una brecha ssmica madura.
La colisin entre las dos placas y el estado en el ciclo ssmico, sugieren que la zona
se encuentra en un rgimen compresivo. Sin embargo sorprende que la solucin del
mecanismo focal para el sismo de Coyuca corresponda a una falla asociada a un rgimen
extensivo.
Algunos autores (P. ej. Singh y Pardo, 1993) han propuesto que la placa cabalgante
de Norteamrica se encuentra en un rgimen tensional, pero hasta ahora la evidencia
provena de sismos pequeos continentales lejanos a la costa. Dada su posicin con
respecto a la zona acoplada (justo encima), este temblor es probablemente el nico evento
de esta naturaleza y magnitud reportado hasta ahora, no solamente a escala local sino
tambin a escala global.
31
Por otro lado, el sismo de Coyuca gener un gran nmero de rplicas, algunas de las
cuales pudieron ser bien localizadas y permitieron restringir adecuadamente el plano de
falla. En este trabajo se presenta el anlisis de algunos aspectos relevantes de la fuente
ssmica del temblor mencionado, as como de la distribucin de sus rplicas.
Marco tectnico
La figura 1 muestra los principales rasgos tectnicos de la regin central del estado
de Guerrero. En la figura se observa que la denominada brecha ssmica de Guerrero est
acotada al noroeste por el rea de ruptura del sismo del 14 de Marzo de 1979 (Valds et al.,
1981) y al sureste por la zona de ruptura del sismo del 14 de Septiembre de 1995
(Mw=7.3).
Las zonas de ruptura de los sismos reportados anteriormente muestran que la
brecha ssmica podra romper en mltiples eventos de magnitudes en el rango de 7.5-7.8.
Sin embargo, la dimensin total de indica que la regin puede romper en un gran evento de
magnitud hasta M=8.2 (Singh y Mortera, 1991). En la figura 1, tambin se muestra la
posicin y el mecanismo focal del sismo de Coyuca (que se discute en una seccin
posterior), localizado justo encima de la brecha ssmica. La figura 2 muestra un esquema de
una seccin perpendicular a la trinchera donde se puede apreciar que, mientras que el sismo
no tuvo una profundidad mayor a 10 Km, la interfase entre las placas de Cocos y
Norteamrica se encuentra a unos 20 Km de profundidad. Por otro lado, se muestra la
posicin relativa (proyectada) del sismo (de trinchera) del 18 de Abril de 2002 (Iglesias et
al., 2003) y de un deslizamiento assmico registrado por estaciones GPS (Kostoglodov,
2003). Las zonas marcadas como ~0% y ~100% acopladas y la regin nombrada como de
transicin fueron trazadas con base en los resultados de una inversin del deslizamiento
assmico para el evento de 2002, a partir de las observaciones registradas en estaciones
GPS (Iglesias et al., 2004).
32
Figura 1. Rasgos tectnicos de la regin central de Guerrero (ver leyenda en la figura). Con lnea
discontinua se muestra la posicin de la trinchera mesoamericana (MAT).
33
34
Hora
2001-10-08
Fuente
Lat.
Lon.
Prof.
Azimut
Echado
Rake
Mw
03:39:20
Regional
17.003
-100.095
8.0
281/89
49/42
-82/-99
5.8
03:39:26
CMT
17.320
-99.890
15.0
263/105
39/53
-107/-77
5.8
2001-10-08
07:16:00
Regional
17.067
-100.095
5.0
257/99
55/37
-103/-72
4.5
2001-10-09
00:34:22
Regional
17.062
-100.088
8.0
274/97
51/39
-92/-88
4.4
2001-10-29
05:23:12
Regional
17.100
-100.135
5.0
263/95
39/52
-99/-83
5.0
05:23:18
CMT
17.490
-99.440
15.0
240/91
44/51
-116/-67
5.0
06:41:37
Regional
17.058
-100.161
8.0
258/61
54/37
-80/-103
4.6
2001-11-23
35
En este trabajo, dedicado al estudio del sismo de Coyuca, el plano de falla fue
dividido en una malla de 6 x 8 subfallas de 4 Km2 de rea. De esta forma, estudiamos un
plano de falla de 12 Km a lo largo del rumbo y 16 Km a lo largo del echado. La orientacin
del plano de falla fue estimada a travs de la inversin del tensor de momentos empleando
datos locales y regionales (=281, 89; =49, 42, =82,99). La distribucin de las rplicas
(que se discutir adelante) muestra claramente que el plano de falla buza hacia el sur. Por
esta razn, se eligi la solucin correspondiente (=89, =42, =-99) para la inversin.
Se utilizaron los registros de la estacin CAIG de banda ancha del SSN; de las
estaciones COYC, OCLL, ATYC, VNTA y ACAP de la red de acelergrafos operados por
el Instituto de Ingeniera (I. I.). Estos acelerogramas fueron integrados numricamente dos
veces, para obtener desplazamientos, y filtrados en una banda de frecuencias entre 0.1 y 0.5
Hz. Como modelo inicial se asign un deslizamiento homogneo de 0.1m. Los lmites del
tiempo de ruptura fueron calculados empleando como velocidades de propagacin extremas
2.2 y 3.2 Km/s.
La figura 4 muestra la ubicacin de las estaciones, as como la proyeccin
horizontal del plano de falla utilizado para la inversin.
36
cm
10
20
30
segundos
40
Figura 5. Ajustes obtenidos a travs de la inversin cinemtica. En lnea continua roja se muestran
los sismogramas observados y en lnea discontinua azul los sintticos.
37
Hipocentro
38
Figura 7 Localizacin (SSN) de las rplicas del sismo del 08 de Octubre de 2001(cuadros rojos). Tambin se
muestra la ubicacin de las estaciones porttiles colocadas despus del sismo de Coyuca (tringulos azul
marino) y las estaciones del I.I. y del SSN (azul claro, invertido y normal, respectivamente).La localizacin
del sismo de Coyuca es mostrada con una estrella verde.
Figura 8 Localizacin de las rplicas del sismo del 08 de Octubre de 2001 utilizando los datos
de la red de estaciones porttiles
39
donde:
m ij = ( dx ij , dyij , dz ij , dij ) es la diferencia entre los parmetros del sismo i con
40
muestra en la figura 10, donde se presentan las localizaciones de las rplicas proyectadas
sobre la seccin A-A.
41
Discusin y Conclusiones
La inversin cinemtica y la solucin del tensor de momentos muestran que el
principal deslizamiento ocurri a una profundidad somera (5-8 Km). Esto sita al sismo de
Coyuca dentro de la placa continental de Norteamrica. Las rplicas registradas en la red
temporal de estaciones de banda ancha muestran claramente que el plano de la falla buza
esencialmente hacia el sur. El nmero de rplicas generadas por el sismo de Coyuca fue
especialmente grande con respecto de la cantidad de rplicas que generan los sismos de
subduccin en Mxico. Un anlisis detallado de la distribucin espacial y temporal de las
rplicas ser necesario para conocer si hubo alguna migracin de la sismicidad o algn
cambio sustancial en los mecanismos focales.
Existe suficiente evidencia de que la placa de Norteamrica, al menos en el centro y
sur de Mxico, se encuentra en tensin (p. ej. Singh y Pardo, 1993; Pasquar et al., 1988).
Sin embargo, el sismo de Coyuca es quiz la nica evidencia de esfuerzos tensionales
actuando sobre esta placa a menos de 80 Km. de la trinchera. Dos posibles escenarios
podran explicar este estado:
- Retroceso de la trinchera
El fenmeno del retiro o retroceso de trinchera es observado cuando el propio peso de la
placa provoca que sta se hunda y, por lo tanto, aumente el ngulo de subduccin. Al
aumentar este ngulo la lnea de trinchera retrocede. En una zona acoplada, el
desplazamiento relativo de la lnea de trinchera provoca extensin en la placa cabalgante.
42
Las condiciones ideales para que ocurra este fenmeno incluyen una placa subduciendo con
un ngulo pronunciado. Sin embargo, esta solucin es poco plausible en este caso ya que la
zona de subduccin mexicana se caracteriza por una subduccin de muy bajo ngulo
(subhorizontal).
- Erosin tectnica.
El proceso de subduccin implica un desgaste en la zona de contacto. La placa que
subduce provoca que parte del material de la zona de contacto, perteneciente a la placa
cabalgante, sea desgastado y acarreado a mayor profundidad. Esto induce un aumento en la
pendiente del arco frontal cercano a la trinchera. La deficiencia de material en esta zona se
traduce en un sistema de fallas normales y en un retroceso del frente de la placa continental.
Parece no haber ejemplos de sismos recientes de fallamiento normal en la placa cabalgante
debido a este proceso. Sin embargo, existen estudios en el norte de Chile (Delouis et al.,
1998) y en Centroamrica (Ranero y von-Huene, 2000; Vanneucchi et al., 2001) donde han
sido observadas fallas normales, no necesariamente activas, en la placa cabalgante, que han
sido atribuidas al proceso de erosin tectnica. Un punto en contra de esta hiptesis, en el
caso de la zona sur-centro de Mxico, estriba en que: s la placa subducida logra deteriorar
la placa cabalgante, sera entonces esperable encontrar un prisma de acrecin importante en
la trinchera. Sin embargo, en el caso de la zona central de Mxico no parece existir un
prisma de acrecin importante.
Una tercera posibilidad, an no estudiada, es que el alto grado de acoplamiento
entre las placas de Cocos y Norteamrica provoque lbulos de esfuerzo tensional en zonas
localizadas en los lmites de la zona sellada.
En primera instancia, ninguna de las tres posibilidades puede ser en descartada. Sin
embargo, no existe ms evidencia de sismos de fallamiento normal, al menos recientes,
localizados en la placa cabalgante sobre la zona de contacto fuertemente acoplada en la
brecha de Guerrero.
Por otro lado, no se puede perder de vista el deslizamiento assmico registrado en la
zona de la brecha de Guerrero. El inici del denominado sismo lento, a finales del 2001,
guarda gran coincidencia temporal con el sismo de Coyuca. Es probable que su principal
deslizamiento est localizado justo debajo de la zona acoplada (Iglesias et al., 2004). Si
43
bien es difcil establecer una correspondencia obvia entre ambos eventos, la coincidencia
temporal y espacial induce a establecer una posible relacin (Kostoglodov et al., 2003).
La presencia del sismo de Coyuca indica la posibilidad de la existencia de fallas del
mismo tipo distribuidas en las zonas aledaas. De esta manera, no se puede descartar que
un sismo de la misma naturaleza y magnitud se localizara muy cerca del puerto de
Acapulco y afectara algunas construcciones de esta ciudad.
Referencias
Bouchon, M. (1982). The complete synthetics of crustal seismic phases at regional
distances, J. Geophys. Res. 87, 1735-1741.
Cotton, F. y M. Campillo (1995), inversion of strong ground motion in the frequency
domain, J. Geophys. Res. 100, 3961-3975.
Hernandez, B., N., Shapiro, S.K. Singh, J.F. Pacheco, F. Cotton, M. Campillo, A. Iglesias,
V. Cruz, J. M. Gmez, and L. Alcntara (2001). Rupture history of September 30, 1999
intraplate earthquake of Oaxaca, Mexico (Mw=7.5) from inversion of strong-motion data,
Geophys. Res. Lett., 28, 363-366
Iglesias, A., S.K. Singh, J.F. Pacheco y M. Ordaz (2002).A Source and Wave Propagation
Study of the Copalillo, Mexico Earthquake of July 21, 2000 (Mw=5.9): Implications for
Seismic Hazard in Mexico City from Inslab Earthquakes, Bull. Seism. Soc. Am., 92, 2,
1060-1071.
Iglesias A., S.K. Singh, J.F. Pacheco, L. Alcntara, M. Ortiz y M. Ordaz (2003). NearTrench Mexican Earthquakes Have Anomalously Low Peak Accelerations, Bull. Seism.
Soc. Am., 93, 3, 953-959,.
Iglesias A., S.K. Singh, A. Lowry, M. Santoyo, V. Kostoglodov, K. M. Larson y S.I.
Franco-Snchez . The silent earthquake of 2002 in the Guerrero seismic gap, Mexico
(Mw=7.6): inversion of slip on the plate interface and some implications, Geof. Int., 43, 3,
309-317
Kostoglodov, V., Larson K. M., S. K. Singh, A. Lowry, J. A. Santiago, S. I. Franco y R.
Bilham (2003). A large silent earthquake in the Guerrero seismic gap, Mexico, Geophys.
Res. Lett. 30, 1807.
44
Nishenko, S.P. y S.K. Singh (1987). Conditional probabilities for the recurrence of large
and great interplate earthquakes along the mexican subduction zone, Bull. Seism. Soc. Am.,
77, 2095-2114.
Ortiz, M., S. K. Singh, V. Kostoglodov y J. Pacheco (2000). Source areas of the AcapulcoSan Marcos, Mexico earthquakes of 1962 (M 7.1; 7.0) and 1957 (M 7.7), as constrained by
tsunami and uplift records, Geof. Int., 39, 337-348.
Pacheco, J.F. y S.K. Singh (1998). Source parameters of two moderate earthquakes
estimated from a single-station, near-source recording, and from MT inversion of regional
data: a comparison of results, Geof. Intern., 37, 95-102, 1998.
Pasquar`e, G., V. H. Garduo, A. Tibaldi, y M. Ferrari (1988). Stress pattern evolution in
the central sector of the MexicanVolcanic Belt, Tectonophys., 146, 352364, 1988.
Randall, G. R., C. J. Ammon y T. J. Owens, Moment-tensor estimation using regional
seismograms from a Tibetan Plateau portable network deployment, Geophys. Res. Lett., 22,
1665-1668, 1995.
Ranero, C. y R. von Huene (2000). Subduction erosion along the Middle America
Convergent Margin, Nature, 404, 13, 748-752.
45
Captulo III
46
ABSTRACT
We invert GPS position data to map the slip on the plate interface during an aseismic, slow-slip event, which occurred in
2002 in the Guerrero seismic gap of the Mexican subduction zone, lasted for ~4 months, and was detected by 7 continuous GPS
receivers located over an area of ~550x250 km2. Our best model, under physically reasonable constraints, shows that the slow slip
occurred on the transition zone at a distance range of 100 to 170 km from the trench. The average slip was about 22.5 cm (Mo~2.97
x1027 dyne-cm, Mw=7.6). This model implies an increased shear stress at the bottom of the locked, seismogenic part of the
interface which lies updip from the transition zone, and, hence, an enhanced seismic hazard. The results from other similar
subduction zones also favor this model. However, we cannot rule out an alternative model that requires slow slip to invade the
seismogenic zone as well. A definitive answer to this critical issue would require more GPS stations and long-term monitoring.
KEY WORDS: Slow earthquake, Guerrero seismic gap, GPS.
INTRODUCTION
Recent continuous geodetic observations, made possible by widespread deployment of GPS receivers, have revealed that slow slip events or silent earthquakes on plate
interfaces are a relatively common phenomenon (e.g. Heki
et al., 1997; Dragert et al., 2001; Ozawa et al., 2001; Miller
et al., 2002). Such observations promise to revolutionize our
understanding of the earthquake source process, interface
coupling, the earthquake cycle, and the rheology of the plate
interface.
Two silent earthquakes have been reported in the
Guerrero seismic gap (Lowry et al., 2001; Kostoglodov et
309
47
A. Iglesias et al.
May 2002, the motion resumed its pre-January 2002 direction at all sites. As an example, Figure 2 shows the time
series of position for station CAYA during the interval January 1997 July 2002. Silent events, one at the beginning of
1998 and another at the beginning of 2002, are highlighted
in the figure.
To determine the change in the position of the sites,
we took the coordinate time series relative to MDO, subtracted the difference in NUVEL1a velocities of the North
American plate at MDO and each site, and inverted, via
weighted least squares, for a best-fit line superimposed by a
hyperbolic tangent function. Table 1 and Figure 1 summarize the resulting estimates of steady-state velocity and
anomalous displacement. The motion during the slow slip
was not perfectly opposite to that during the steady-state
phase, instead it had a significantly less strike-parallel component of motion.
INVERSION FOR SLIP
To invert for slip on the plate interface, we use the fault
geometry shown in Figure 1. The strike of the fault is chosen to coincide with the middle America trench (azimuth
=289). The length of the fault along strike is 600 km. The
horizontal projection of the fault has a width of 350 km (Figure 1). The fault is subdivided into 3x4 elements, three along
strike and four in the down-dip direction. A larger number
of elements is not warranted in view of the small number of
observations. We denote the elements by (i,j), i=1 to 3 and
j=1 to 4. The elements (2,j) coincide with the Guerrero seismic gap. We note that the widths of the elements along the
down-dip direction have been chosen to reflect our current
knowledge of the seismic behavior of the plate interface
along the Mexican subduction zone. Quite generally, large,
shallow thrust earthquakes in Mexico do not occur between
the trench and a distance of ~50 km towards the coast. The
seismogenic zone roughly extends from 50 km to 100 km,
and the transition zone extends beyond 100 km. The element widths have been chosen to reflect this (Figure 1): 050 km; 50-100 km; 100-170 km; 170-350 km. Note that the
division of the transition zone in two segments, 100 km to
170 km and 170 km to 350 km, is arbitrary. As Figure 1
shows there are more GPS sites in the Guerrero gap than in
the adjacent regions. For this reason, we expect the slip on
elements (2,j) to be better resolved than those on other elements.
DATA
Kostoglodov et al. (2003) presented the time series of
positions, relative to McDonald Observatory (MDO), Texas;
of each of the seven GPS sites. These data exhibit all sites
moving NE before January 2002. At this time, the sense of
motion reversed and continued to do so for ~4 months. In
310
In the inversion, the displacement from each rectangular element is calculated using closed form expressions developed by Okada (1992). The rake of the slip vector, , is
taken as an unknown parameter but is assumed to be the
same for each element. The displacement from slip over the
elements of the fault plane may be written as:
48
Fig. 1. (Top) Location of permanent GPS sites in and near the Guerrero seismic gap, Mexico. The vectors at the sites indicate horizontal
position change for a year of interseismic, steady state strain accumulation phase (dark arrows) and during the January-April, 2002 slow-slip
(white arrows). The vectors near the mid-America trench (MAT) illustrate velocity of Cocos plate relative to NOAM plate. Large rectangle is
the horizontal projection of the plate interface over which the slip was inverted from the GPS data. The rectangle is divided in (3x4) elements.
The element number is shown in parentheses. Guerrero gap coincides with elements (2,j). (Bottom) Idealized geometry of the Benioff zone
along AA used in the inversion. The white and the dark segments of the interface indicate the element widths.
49
311
A. Iglesias et al.
Table 1
Position change of GPS sites in southern Mexico relative to McDonald Observatory, Texas. Z is positive upward. is the
standard deviation.
Site
Steady-state phase
(yearly)
CAYA
ACAP
IGUA
YAIG
ZIHP
PINO
OAXA
uk =
NN
(cm)
EE
(cm)
Zz
(cm)
NN
(cm)
EE
(cm)
Zz
(cm)
1.640.01
1.960.02
1.540.05
0.830.02
2.020.04
2.130.06
2.240.09
1.600.02
1.710.03
0.990.08
0.770.03
1.890.06
1.920.08
1.630.13
1.040.06
1.500.10
-0.470.23
0.770.08
-0.220.18
1.140.24
-1.580.38
-5.570.06
-4.490.05
-4.350.07
-1.970.05
-2.040.06
-2.500.09
-2.200.12
-1.760.09
-0.960.07
-1.180.11
-1.470.10
-1.330.09
-0.600.13
-2.340.18
-6.120.27
-2.050.22
1.980.34
2.900.26
-6.000.26
-7.200.38
3.280.50
(Gd
k
k
i,j * Si,j * sin( ) +Gs i,j * cos( ))
i=1 j=1
misfit =
(u
L =1
obs
L
u Lpre
* wL ,
where n is the number of data points (n=number of components multiplied by number of stations), u obs
, u Lpre are the
L
312
Slow-event phase
Jan-Apr 2002 (4 months)
observed and predicted displacements for the L-th component-station and wL (weight for each component-station) is
the reciprocal of standard deviation over data () .
Slow-slip phase
The data set consists of 21 values. The unknown parameters are 13, corresponding to 12 slip amplitudes on (3x4)
elements and 1 slip direction. First we inverted for the slip
distribution without any further constraint. The results showed
a large slip on the element (2,3) that lies between 100 and
170 km from the trench. The slip on this element was at least
twice greater than on any other element. We then performed
several tests by changing the number of elements in the downdip direction. Basically, the solutions showed that the slip on
the central strip (2,j) (corresponding to the Guerrero seismic
gap) is well resolved. The slip on the elements (1,j) and (3,j),
however, significantly changed with any change in the number of the elements, indicating that the slip distribution on
these elements is not well resolved from the available data.
These initial tests suggest that the solutions are unstable because of a large number of parameters (13), relatively small
number of data points (21), and their spatial distribution.
Before performing additional inversions, we imposed
further constraints that are physically reasonable. We required
that the slip on the elements (i,1) be equal. The same constraint was imposed on the elements (i,4). These bands represent the shallowest and deepest portion of our model (Figure 1). In the initial tests, the lateral elements (1,2), (1,3),
(3,2), (3,3) had shown significant instability due to lack of
stations. For this reason, we constructed bigger blocks by
50
Fig. 2. Time series of the position of CAYA station for the interval from January 1997 to July 2003. Shaded rectangles show the two silent
events recorded by the station. (Modified from Kostoglodov et al., 2003)
model (10.8 cm) is similar to that from the model where slip
on each of the 13 elements is free (7.9 cm). The total seismic
moment release, assuming a rigidity =3.5x1011 dyne/cm2,
is 2.97 x1027 dyne-cm (Mw7.6).
Figure 3 (top) shows that the slow slip was mostly confined to the element (2,3). This element extends from 100 to
51
313
A. Iglesias et al.
Fig. 3. (Top) Slip distribution on the plate interface obtained from inversion of the GPS data during the slow-slip phase (January-April, 2002;
4 months). The area in the figure corresponds to the rectangle shown in Figure 1. The numbers in parentheses indicate elements as in Figure 1
(top). Elements separated by dashed lines have been merged. Vectors shown by continuous and dashed lines indicate observed and calculated
displacements, respectively. For clarity, vertical components are shown to the right side of the figure. (Middle) Same as top but with elements
(2,2) and (2,3) merged together. (Bottom) Same as middle but for the steady-state velocities.
314
52
Steady-state phase
The data on yearly position change of GPS sites during
the steady-state phase of the deformation (Table 1) were inverted using the same fault geometry and constraints as in
the previous case. The slip distribution, shown in Figure 3
(bottom), suggests that in the Guerrero gap (elements (2,j))
the data can be explained by an average back slip (20) of
~4.1 cm on the plate interface between 50 and 170 km from
the trench (elements (2,2) and (2,3)). The plate interface further down dip, in the distance range of 170 to 350 km from
the trench (element (2,4)), requires a back slip of ~1.6 cm.
The back slip in this element appears to be real in view of
the position change of inland sites of IGUA and YAIG (Figure 1). The inversion also shows that the elements nearest to
the trench (i,1) require a back slip of ~1.9 cm. The back slip
on the lateral elements (1,2) and (1,3), and (3,2) and (3,3)
are ~ 6 cm and 5 cm, respectively. The direction of back slip
(rake) is 108, which is consistent with the direction of convergence of Cocos with respect to NOAM of 35.
DISCUSSION AND CONCLUSIONS
The displacement vectors at GPS sites above the subduction zone of Guerrero, Mexico, during the steady-state
phase of strain accumulation are in agreement with the relative convergence of Cocos with respect to NOAM (5.6 cm/
yr towards N35E). The inversion of the GPS data, with
physically reasonable constraints, shows an almost completely locked plate interface in the distance range of 50 to
170 km from the trench and a partially locked (~35%) interface further down dip between 170 and 350 km from the
trench (Figure 3, bottom). The direction of the back slip vector, -108, is consistent with the direction of the relative convergence vector.
The reversed motion of the GPS sites, which began in
January 2002 and lasted for around four months, demonstrates
the occurrence of a large silent earthquake. Our best model
shows that the slow slip occurred below the Guerrero seismic gap on the plate interface that extends from ~100 to 170
km (slip ~22.5 cm) and from 170 to 350 km (slip ~3.5 cm).
The direction of slow slip was 91. In this model, the slow
slip did not extend over the upper, locked portion of plate
interface (~50 to 100 km from the trench). Thus the slip of
22.5 cm in four months cancelled the strain accumulated on
the element (2,3) (~100 to 170 km from the trench; Figure 3,
top) during about five years of steady-state loading. This is
our preferred model. A consequence of this model is an increased shear stress at the bottom of the locked interface,
thus enhancing the probability of rupture of the Guerrero
seismic gap in the near future. Support for this model comes
from slow slip on the transition zone reported in other regions where the age and relative speed of the subducting plate
is similar, e.g., (Draggert et al., 2001) and before 1944
53
315
A. Iglesias et al.
small events using empirical Green functions and simulated annealing. Geophys. J. Int., 125, 768-780.
54
SAVAGE, J. C., 1983. A dislocation model of strain accumulation and release at a subduction zone. J. Geophys.
Res., 88, 4984-4996.
SINGH, S. K., J. HAVSKOV and L. ASTIZ, 1981. Seismic
gaps and recurrence periods of large earthquakes along
the Mexican subduction zone. Bull. Seism. Soc. Am., 71,
827 843.
SINGH, S. K. and F. MORTERA, 1991. Source-time functions of large Mexican subduction earthquakes, morphol-
55
317
Captulo IV
56
Bulletin of the Seismological Society of America, Vol. 93, No. 2, pp. 953959, April 2003
Abstract
It has previously been reported that regional seismograms of earthquakes that occur near the Middle America trench are relatively deficient at high
frequencies. Based on this observation, an algorithm has been proposed for detecting
potentially tsunamigenic earthquakes and issuing tsunami alerts. It is reasonable to
expect relatively low peak accelerations during these earthquakes. In this note, we
present evidence that this is indeed the case. This explains why the seismic alert
system for Mexico City, with sensors located along the coast, does not trigger during
some earthquakes. Low peak accelerations from near-trench earthquakes also have
important implications in seismic hazard estimation.
Introduction
Shapiro et al. (1998) showed that earthquakes in Mexico
that occur near the Middle America trench are abnormally
depleted in high-frequency radiation at the broadband station
CUIG, located in Ciudad Universitaria, Mexico City. This
station lies about 300 km from the nearest point of the Pacific
coast (Fig. 1). Shapiro et al. suggested that the converse may
also be true, that is, all events depleted in high-frequency
radiation at CUIG may be located near the trench. Because
of their near-trench location, these earthquakes, if they are
large enough in magnitude, are potentially tsunamigenic. Indeed, on the basis of these observations, Shapiro et al. proposed a fast and simple method for identifying tsunamigenic
earthquakes along the Mexican subduction zone. The
method relies on the ratio of the total radiated energy to
the high-frequency energy (f 1 Hz), ER, computed from
the CUIG seismograms. The same seismograms are also used
to compute preliminary magnitudes, Ma and Me, of Mexican
earthquakes (Singh and Pacheco, 1994). Ma is based on the
amplitude of long-period (1530 sec) waves, while Me is
based on seismic energy computed from the seismograms at
CUIG. These magnitudes are related to Mw. The estimation
of magnitude and ER can be accomplished in about 5 min
after the origin of the earthquake, and a useful tsunami alert
can be issued. A precise location of the event is not required;
an ER greater than 100 is found to be sufficient evidence that
the earthquake is near trench.
Since the near-trench events are depleted in highfrequency radiation, it follows that during such earthquakes
the peak accelerations, Amax, at CUIG and other stations in
the Valley of Mexico would be less than during near-coast
earthquakes of the same magnitude. It also seems reasonable
to expect relatively low Amax along the coast for near-trench
earthquakes. Unfortunately, the issue of Amax from neartrench earthquakes was not addressed in the previous article.
57
954
Short Notes
sites and, hence, are less likely to cause damage from strong
ground motions. However, there is catch to this good news.
Because of its low acceleration, the earthquake may go unfelt by the coastal inhabitants and a large tsunami may be
their first exposure to the phenomenon. This was the case
during the 2 September 1992 tsunami earthquake of Nicaragua (mb 5.3; Ms 7.2; Mw 7.6). The ground motion during
the earthquake was very weak along the entire coast of Nicaragua. The earthquake was lightly felt, if at all, in the
Table 1
Source Parameters of the Earthquake of 18 April 2002 and Some Its Aftershocks
Date
(yymmdd)
020418*
020418
020418
020418
020418*
Time
05:02:43.5
07:24:27.3
08:01:34.9
08:28:05.7
17:57:23.9
Latitude
(N)
Longitude
(E)
16.75
16.77
16.82
16.83
16.96
101.06
101.00
101.01
101.07
101.40
Depth
(km)
6.0
2.0
2.0
2.0
4.0
Strike
Dip
Rake
M0
(dyne cm)
291
89
1.5 10
2.4 1022
273
17
81
8.8 1024
26
6.7 Mw
4.2 Mw
4.3 Mc
4.3 Mc
5.9 Mw
All epicentral locations and depths are from local/regional data. Depths are poorly constrained.
58
955
Short Notes
100oW
105 W
95oW
North American
Plate
20oN
20oN
CUIG
9
Latitude
8 10
15
2 14
21 20
MA
Cocos Plate
T
3 7 16
4
17 11
5
12 13
15 N
15 N
100oW
105 W
95oW
Longitude
Table 2
Earthquakes and Their Source Parameters
No.
Date
(yy.mm.dd)
Latitude
(N)
Longitude
(E)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
85.09.19
85.09.21
89.04.25
93.10.24
93.11.13
94.12.10
95.09.14
95.10.06
95.10.09
95.10.12
96.02.25
96.02.26
96.03.19
96.07.15
97.01.16
97.01.21
97.07.19
96.02.21
96.11.12
02.04.18
02.04.18
18.1
17.6
16.6
16.5
15.7
18.0
17.0
18.8
18.8
18.7
15.6
15.7
15.5
17.5
18.1
16.4
16.0
9.6
15.0
16.8
17.0
102.7
101.8
99.5
99.0
99.0
101.6
99.0
104.5
104.5
104.2
98.3
98.2
97.6
101.1
102.9
98.2
98.2
79.6
75.7
101.1
101.4
M0
(dyne cm)
Mw
ER
8.0
7.5
6.9
6.6
5.7
6.4
7.3
5.8
8.0
5.9
7.1
5.5
5.8
6.6
5.5
5.5
6.7
7.5
7.7
6.7
5.9
45.9
35.8
53.1
31.5
173
19.5
62.4
40.8
258
46.6
174
472
160
7.8
20.2
5.6
227
1180
10.1
519
132
1.1
2.9
2.4
1.0
4.1
5.2
1.3
5.8
1.1
1.0
5.5
2.3
6.9
1.0
2.2
2.1
1.2
2.2
4.6
1.5
8.8
1028
1027
1026
1026
1024
1025
1027
1024
1028
1025
1026
1024
1024
1026
1024
1024
1026
1027
1027
1026
1024
Es*
(erg)
8.5
2.2
2.5
1.1
2.3
2.8
1.2
1.8
3.0
1.3
3.3
2.3
5.7
2.2
Es/M0
1022
1022
1021
1021
1019
1020
1022
7.7
7.6
1.0
1.1
5.6
5.4
9.2
1022
1019
1021
1.6
3.0
2.4
1020
3.3
1020
1021
1022
1.9
2.6
4.8
106
106
105
105
106
106
106
106
106
106
106
106
106
106
V ( f )df
2
ER
(1)
V ( f )df
2
59
956
Short Notes
(2)
60
957
Short Notes
18
1000
20
12
17
13
11
21
100
near-trench events
near-coast events
7
10
10
19
14
16
5.0
6.0
7.0
8.0
Magnitude, Mw
10
-5
Es/Mo
6
5
19
14
10
11
18
17
9
10
-6
101
(3)
4
15
102
103
Are the relatively depleted high-frequency radiation observed at CUIG and the low accelerations observed along the
coast and in the Valley of Mexico during near-trench earthquakes a source or a path effect? The large disparity between
Ms and Mw of tsunami earthquakes, which occur near the
trenches (Kanamori and Kikuchi, 1993), points to a source
effect, as do the relatively larger rupture duration (Bilek and
Lay, 1999) and the relationship between ER and Es /M0 of
shallow near-trench earthquakes. On the other hand, the seismic waves from these shallow earthquakes must suffer a
strong path effect as they get trapped in the low-velocity
accretionary prism (see, e.g., Shapiro et al., 2000) and propagate in shallow oceanic layers to coastal and inland stations
(Fig. 3). The complexity of seismograms of near-trench
Mexican earthquakes (Fig. 3) makes it difficult to study the
details of source characteristics. To study the source of the
18 April mainshock, we deconvolved the mainshock seismograms recorded at the broadband stations of ZIIG, CAIG,
and PLIG (Fig. 1) by those of the aftershocks, thus minimizing the path effect. We still could not obtain a stable source
time function, most probably because the aftershocks were
not appropriate empirical Greens functions. While the most
likely cause of depleted high-frequency radiation and low
accelerations during near-trench earthquake is the source, we
cannot rule out that the path plays an important role as well.
Observed Tsunami during the 18 April 2002
Mainshock
We have provided evidence that the 18 April mainshock
was potentially tsunamigenic. In fact, a small-amplitude tsunami was recorded at Acapulco and Zihuatanejo. At these
sites, digital sensors take one pressure reading each second,
and the average value of 6 min of data is saved. In Figure
6, which shows the residual tide, a tsunami, whose amplitude
is about twice the background noise, can be seen at both
stations. As these sensors were not installed during previous
Mw 6.5 earthquakes in the region, it is not possible to assert
that the tsunami during the 18 April event is anomalously
large.
Conclusions
It has been previously reported that the earthquakes that
occur near the Middle America trench are depleted in high-
61
958
Short Notes
Figure 5. Amax versus distance for four near-trench events. The curves depict the
predicted Amax from near-coast earthquakes. Crosses mark recorded Amax. The horizontal lines for the 18 April 2002 earthquakes indicate the range of distance over which
Amax was less than 3 Gal.
Acknowledgments
We are grateful to the technical staff that maintains the broadband
seismological stations (especially Jorge Estrada, Jesus Perez Santana, and
Jose Lus Cruz) and the strong-motion networks (mainly David Almora,
Miguel Torres, Ricardo Vasquez, Juan Manuel Velasco, and Mauricio Ayala). Nicolas Shapiro kindly revised the manuscript. We thank Emile Okal
for his thoughtful comments. The tsunami data was provided by the sea
level network of Secretara de Marina de Mexico. The research was partially supported by DGAPA UNAM Project IN 111601.
62
959
Short Notes
Figure 6.
Residual tide at Acapulco and Zihuatanejo. Crosses indicate data (one sample per 6 min).
The light curve shows the envelope. The zero along
the time axis corresponds to the origin time of the
mainshock of 18 April 2002 (Table 1).
References
Anderson, J. G., J. N. Brune, J. Prince, R. Quaas, S. K. Singh, D. Almora,
P. Bodin, M. Onate, R. Vasquez, and J. M. Velasco (1994). The Guerrero accelerograph network, Geofis. Int. 33, 341371.
Bilek, S. L., and T. Lay (1999). Rigidity variations with depth along interplate megathrust faults in subduction zone, Nature 400, 443446.
Boatwright, J., and G. L. Choy (1986). Teleseismic estimates of the energy
radiated by shallow earthquakes, J. Geophys. Res. 91, 20952112.
Espinosa-Aranda, J. M., A. Jimenez, G. Ibarrola, F. Alcantar, A. Aguilar,
M. Inostroza, and S. Maldonado (1995). Mexico City seismic alert
system, Seism. Res. Lett. 66, 4253.
Ide, S., F. Imamura, Y. Yoshida, and K. Abe (1993). Source characteristics
of the Nicaraguan tsunami earthquake of September 2, 1992, Geophys. Res. Lett. 9, 863866.
Kanamori, H., and M. Kikuchi (1993). The 1992 Nicaragua earthquake: a
slow tsunami earthquake associated with subducted sediments, Nature
361, 714716.
Montalvo-Arrieta, J. C., E. Reinoso, F. J. Sanchez-Sesma, S. K. Singh,
J. Pacheco, and M. Ordaz (2003). The seismic response of the hill
zone in Mexico City: a review and new findings (in preparation).
Newman, A. V., and E. A. Okal (1998). Teleseismic estimate of radiated
seismic energy: the Es /M0 discriminant for tsunami earthquakes,
J. Geophys. Res. 103, 26,88526,898.
Okal, E. A., and A. V. Newman (2001). Tsunami earthquakes: the quest
for a regional signal, Phys. Earth Planet. Interiors 124, 4570.
Ordaz, M., and S. K. Singh (1992). Source spectra and spectral attenuation
of seismic waves from Mexican earthquakes, and evidence of amplification in the hill zone of Mexico City, Bull. Seism. Soc. Am. 82,
2443.
Ordaz, M., J. M. Jara, and S. K. Singh (1989). Riesgo ssmico y espectros
de diseno en el estado de Guerrero, in Mem VIII Congreso Natl. de
Ingeniera Ssmica y VII Congreso Nac. de Ingeniera Estructural 2,
D40D56, Acapulco.
Shapiro, N. M., K. B. Olsen, and S. K. Singh (2000). Wave-guide effects
in subduction zones: evidence from three-dimensional modeling,
Geophys. Res. Lett. 27, 433436.
Shapiro, N. M., S. K. Singh, and J. Pacheco (1998). A fast and simple
diagnostic method for identifying tsunamigenic earthquakes, Geophys. Res. Lett. 25, 39113914.
Singh, S. K., and M. Ordaz (1994). Seismic energy release in Mexican
subduction zone earthquakes, Bull. Seism. Soc. Am. 84, 15331550.
Singh, S. K., and J. Pacheco (1994). Magnitude of Mexican earthquakes,
Geofis. Int. 33, 189198.
Venkataraman, A. (2002). Investigating the mechanics of earthquakes using
macroscopic seismic parameters, Ph.D. Thesis, California Institute of
Technology, Pasadena.
Appendix
Note Added in Proof
G. L. Choy (personal comm., 2003) reports Es 2
1020 ergs for the 18 April 2002 (Mw 6.7) event. Thus Es /M0
1.5 106 for this event. Since ER 519 (Table 2),
this event agrees with our expectation that if ER 100 then
Es /M0 is less than 3 106.
Instituto de Geofsica
Universidad Nacional Autonoma de Mexico, C.U.
04510 Mexico, D.F.
Mexico
(A.I., S.K.S., J.F.P.)
Instituto de Ingeniera
Universidad Nacional Autonoma de Mexico, C.U.
04510 Mexico, D.F.
Mexico
(L.A., M.Ord.)
Departamento Oceanografa Fisica
Centro de Investigacion Cientfica y de Educacion Superior de Ensenada
Ensenada, B.C.
Mexico
(M.Ort.)
Manuscript received 1 August 2002.
63
Captulo V
64
Abstract
The seismic alert system (SAS) for Mexico City, an impressive technological feat, has now been in
operation for more than 10 years. The SAS takes advantage of the fact that the city is located more than 300
km from the foci of potentially damaging earthquakes. The system consists of 15 accelerometers located
along the coast of the State of Guerrero, above a segment of subduction plate boundary that is a mature
seismic gap. An algorithm estimates magnitude of an event from the near-source accelerograms, and issues
public and restricted alerts for M6 and 5M<6, respectively. An evaluation of the SASs performance
during 1991-2004 reveals a surprisingly high failure rate. This poor performance results from an inadequate
detection algorithm and the limited areal coverage by the SAS. These two factors render the alert system of
limited use.
In this paper we propose an alternative strategy for detecting potentially damaging earthquakes to Mexico
City that differs substantially from the one presently implemented by the SAS. It is developed from the
analysis of near-source recordings of Mexican earthquakes since 1985 and the corresponding ground
motions recorded in Mexico City. In our proposed scheme, the alerts are based on the relationship between
root-mean-acceleration (Arms) in the near-source region and the expected Amax at a reference site in
Mexico City, CU. Tests are performed using unfiltered and band-pass filtered (0.2-1.0 Hz) accelerograms.
The choice of the filter corresponds to the frequency band of amplification of seismic waves in the lake-bed
zone. The results suggest that only a single level of general public alert may be the best option. This alert
would be issued when the near-source Arms computed over a window of 10s exceeds 5 gal for unfiltered
records or 1.0 gal for the filtered ones. The corresponding values of expected (Amax)CU are 3.5 and 2.0 gal,
respectively. We find that the use of band-pass filtered accelerograms leads to a lower failure rate of alerts.
The data since 1985 suggests that such an alert, on average, would occur about once a year on average. It
would include most earthquakes felt by most persons in the lake-bed zone and, most of all, not miss any
damaging event. The proposed strategy, along with deployment of about 40 sensors in three concentric
rings centered at Mexico City, would considerably improve the performance of SAS and, potentially, save
thousands of lives.
65
Introduction
The metropolitan area of Mexico City is inhabited by about 20 million persons.
Although the city is located more than 300 km from the Pacific coast where most large
earthquakes originate, it still suffers frequent earthquake disaster. Normally, the
amplitudes of seismic waves at such distances are sufficiently diminished so that even
large earthquakes do not cause any damage. The principal cause of this unexpected
phenomenon is well known: an extraordinary amplification of seismic waves in the
frequency band of 0.2 to 1.0 Hz resulting from the soft clays that underlie the lake-bed
zone of the Valley of Mexico (e.g., Singh et al., 1988a,b). The most recent example was
the Michoacan earthquake of 1985 (Mw8.0) which originated at a distance of about 350
km from the city. The collapse of buildings during the earthquake killed about 10,000 and
injured 30,000 persons.
Relatively large distance (>300 km) between the foci of most of the potentially
damaging earthquakes and the city provides a unique opportunity for a seismic alert
system (SAS). The large amplitude S waves reach Mexico City more than 85 s after the
origin. A quick detection of the occurrence of an earthquake and an estimation of its
damage potential to the city is possible by deploying sensors above the epicentral region.
This can provide about 60 s of alert time to Mexico City and, potentially, save thousands
of lives.
In fact, a SAS for Mexico City has been in operation since August
1991(Espinosa-Aranda et al., 1995; Espinosa-Aranda and Rodriguez, 2003). It is an
impressive technological achievement. The sensors deployed by the SAS presently cover
a 300 km-long segment of the Guerrero coast (Figure 1), a known mature seismic gap
(Singh et al., 1981). It consists of 15 accelerometers located 25 km apart. The system
detects P and S waves at the sensor nearest to the focus, and computes the energy in the
accelerogram over a time window beginning with S wave and lasting twice the (S-P) time
(generally about 6 to 8 s). It also computes the rate of accumulation of energy during this
window. These two parameters are used to estimate the magnitude, M, of the earthquake
(Espinosa-Aranda et al., 1989). The system issues public and restricted alerts for M6
66
and 5M<6, respectively. At least two stations must confirm the occurrence of the event
before the warning is automatically broadcast by SAS.
In this study, we evaluate SASs performance thus far, and, based on the analysis
of strong motion data in Mexico, suggest modifications that may improve its
effectiveness.
Figure 1. Map of Mexico showing, in orange, the area covered by Mexico Citys seismic alert system
(SAS). Triangles and rectangles indicate interplate and non-interplate events. Green and red symbols
show locations of events for which the SAS issued restricted and public alerts, respectively during the
period August 1991-August 2004. Symbols with white interior are events which generated (Amax)CU5 gal.
Gray symbols with white interior: events before the SAS coverage (1964-July 1991) with (Amax)CU5 gal.
Events with (Amax)CU5 gal during the SAS coverage but with no alert are shown by white symbol. The
three concentric semi-circles show the geometry of a proposed array which would provide a more adequate
coverage by the SAS. Sensors would be located about 60 km apart in each of the semi-circular rings (for a
total of about 40).
67
Table 1 summarizes SASs performance since August 1991 when the system
became operational. During this period, SAS issued 46 restricted and 11 public alerts (J.
M. Espinosa-Aranda, personal communication, 2004). Figure 1 shows epicenters of these
earthquakes. As Table 1 illustrates, a restricted alert (5M<6) was justified only in 15 of
the 46 cases, in 27 cases no alert should have been given, and in 4 cases (with M6) a
public alert should been issued. Only three of the 11 public alerts actually fulfilled the
requirement of M6. During this period the SAS also gave one false public alert when no
earthquake occurred and, in one case, it failed to give any alert when it should have
issued a public one. As Table 1 shows, the SASs failure rate is high.
Magnitude estimated
by SAS as the basis
of the alert
Number of
alerts issued
True magnitude
distribution of the events
Restricted
5M<6
46
4M<5
27
Public
M6
11
5M<6
15
6
M6
4
3
In reality, we do not need an estimate of the magnitude for issuing a seismic alert
for Mexico City. The performance of the SAS may be better evaluated by associating
alerts with recorded ground accelerations in Mexico City. For this purpose, we analyze
recordings at a reference station in the city where accelerographs have been in continuous
operation since the mid sixties. This station, CU, is located on basaltic lava flows. The
peak horizontal acceleration, Amax, in the lake-bed zone of the city, where much of the
damage occurs during earthquakes, is 4 to 5 times greater than at CU (Singh et al., 1987,
1988c). We define Amax at CU by
68
( A max)CU = a N2 + a E2 + a Z2
1
2
(1)
Number of alerts
Public Alert
5
4
3
2
1
0
1
9 10 11 12 13 14 15 16 17 18 19
(Amax)CU,gal
Number of alerts
Restricted alert
40
30
20
10
0
1
9 10 11 12 13 14 15 16 17 18 19
(Amax)CU,gal
69
Lat
Long
To develop a more appropriate strategy for issuing reliable seismic alerts for
Mexico City, we analyzed accelerograms recorded at CU and the corresponding nearsource recordings of the same events. Extensive near-source recordings in Mexico began
in 1985 with the installation of the Guerrero Accelerographic Array (Anderson et al.,
1994). Although the free-field accelerographic networks operated by the Institutes of
Engineering and Geophysics, UNAM have steadily grown, the near-source recordings are
not available for many significant earthquakes.
Table 3 lists 45 earthquakes which gave rise to the largest Amax at CU during
1985-2004. The events are listed in descending order of Amax at CU. The table indicates
the event type: shallow thrust-faulting interplate earthquake along the Pacific coast
(10<H<30 km), steeply-dipping thrust or normal-faulting inslab earthquake (35<H<120
km), and shallow local earthquake near Mexico City. We note that six of the first 10
events in the table are inslab, normal-faulting earthquakes. In Mexico, these events occur
below the coast as well as below the altiplano. These earthquakes are known to be
70
In Table 3 we do not assign a number to the local earthquakes and those events
which were recorded at CU but did not produce a near-source recording. For events 18,
28, 30, and 35, the epicentral distance to the closest station, S, was greater than 100 km.
Although we analyze these four events, we do not include them in the statistics. The first
two events in Table 3 devastated Mexico City occurred and the third and fourth events
were strongly felt but resulted in only minor damage to the city. Amax at CU exceeded 8
and 7 gal during the first 7 and 10 events, respectively.
From previous studies, it is well known that the peak acceleration, Amax, in the
near-source region, as a function of M shows large scatter and is, essentially, independent
of M for M greater than about 5.5 (e.g., Singh et al., 1989). Both the source and the site
effects are responsible for the large scatter. It follows that near-source Amax as a measure
of M is unsatisfactory. Since larger earthquakes last longer, such earthquakes cause
sustained high acceleration over longer duration. This suggests that root-mean-square
acceleration, Arms, computed over an adequate time window, td, may be a more useful
measure of the size of the earthquake. The window td should be sufficiently small to
provide as much warning time as possible and, yet, large enough to provide information
about the size of the earthquake.
71
Date
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
Lat
Long
H
(km)
22
22
60
17
50
22
M
8
7.6
6.9
6.9
5.9
7.3
40 7.4
50
40
50
21
21
26
32
27
35
6.4
7.1
6.2
6.9
6
7.5
6.5
6.6
6.6
8
16
15
53
22
21
21
5.9
6
8
6.3
5.8
5.3
6
5.9
69
54
42
36
27
39
64
15
5.2
6.5
5.2
5.1
6.3
5.3
5.9
6.0
5.1
18 5.4
16
6
25
15
5.9
5.5
5.4
6.7
5.9
28 5.5
5.0
29 5.4
17 6.4
168 6.8
4.2
4.6
16 5.6
(Amax)CU
(gal)
42.156
19.669
18.072
17.042
15.381
14.587
11.022
10.932
10.695
7.863
7.687
7.181
7.036
6.614
6.001
5.904
5.042
4.892
4.796
4.353
4.268
4.149
3.960
3.883
3.735
3.631
3.450
3.199
2.830
2.547
2.428
2.298
2.063
2.033
1.968
1.916
1.768
1.699
1.565
1.539
1.520
1.454
1.245
1.208
1.200
1.190
1.185
1.161
1.130
1.110
1.055
1.029
1.012
1.012
s1
(km)
9,98,56
39,89
51,107
23,27
63,76,78
48,74,58
CU2
(km)
402,314,350
307,297
215,115
284,288
142,115
303,285,344
26,96,51
431,367,411
BALC
CALE
COMD
CALE
SLUI
MZ02
CALE
AZIH
COPL
33
35
13
43
18
52
20
31
3
259
401
193
401
291
474
401
307
303
COYC
SMR2
MZ01
AZIH
AZIH
ATYC
AZIH
PNIG
7
163
50
66
32
47
77
19
279
284
543
307
307
270
297
344
TNLP
ZIIG
CHFL
TNLP
HUIG
SMR2
ZIIG
PET2
SLUI
PNTP
31
93
52
151
29
163
89
77
40
25
307
307
214
142
512
284
307
297
291
351
SMR2
SLUI
AZIH
PNTP
126
17
7
62
284
291
297
351
PAPN
JAMI
ATYC
PET2
34
39
30
41
297
367
270
300
UNIO
VNTA
PNTP
37
37
47
314
275
351
Station
Type3
T
T
IS
T
IS
T
L
IS
L
L
IS
IS
IS
T
T
T
IS
T
T
L
N
T
T
T
T
T
T
T
L
IS
IS
T
T
T
T
T
T
T
T
?
IS
T
T
T
?
T
T
T
T
?
?
T
T
T
s: Epicentral distance
CU: Distance between the near-source station and CU
3
?: Unknown. All other symbols are the same as in Table 2
1
2
72
For our analysis of the near-source accelerograms, we compute I(td), Arms(td), and
(td), where
td
I (t d ) = a N2 + a E2 + a Z2 dt
(2)
1 I (t ) 2
Arms(t d ) = d
3 td
and, (t d ) =
I (t d )
I (t d 2 )
(3)
(4)
Figures 3a and 3b illustrate what we may expect from the use filtered records.
They also help understand the results that follow. In Figures 3a, the left frames are nearsource accelerograms (NS component) of 6 selected events from Table 3, and the center
and right frames are the plots of IN(t) using unfiltered and filtered accelerograms,
respectively. Here
t
I N ( t ) = a N2 ( t ) dt
0
The integration in the plots begins with the S-wave arrival and lasts for 10s.
73
Figure 3a. (Left): NS component of near-source accelerograms of 6 events (5.9Mw8.0) from Table 3.
(Middle): plots of
Unfiltered accelerograms are used in computing IN(t).(Right): IN(t) plots using band-passed filtered
(0.2 -1.0 Hz) accelerograms.
Figure 3b. (Left): NS component of near-source accelerograms of the first four events of Table 3. For each
event more than one near-source trace is shown. (Middle). IN(t) plots using unfiltered accelerograms.
(Right): IN(t) plots using band-passed filtered (0.2 -1.0 Hz) accelerograms.
74
Several points may be noted from Figure 3a: (1) Near-source Amax is independent
of M in the magnitude range 5.8 to 8.0. (2) IN(10), computed from unfiltered record, does
not scale with M. The largest IN(10) corresponds to the smallest event 17 (M5.8). (3)
When computed using filtered records, IN(10) of the smaller events (M~5.9) are much
smaller than for the larger events. (4) For M~5.9 events, IN(t) is nearly flat 4 to 5s after
the arrival of S wave and, hence, (10) = IN(10)/ IN(8)~1. For large earthquakes (events
1 and 6), IN(t) keeps increasing up to 10s, thus (10)>1.This suggests that (10) in
Equation 4 could be used to discriminate between small and moderate earthquakes, and
large earthquakes. Unfortunately, (10) for event 4 (M6.9) is about 1, which would
classify it as a small or moderate earthquake. As mentioned earlier, this earthquake was
strongly felt in Mexico City and gave rise to fourth largest Arms at CU (Table 3). Any
useful algorithm must classify this event for a public alert.
Figure 3a may suggest that for earthquakes with multiple near-source recordings,
IN(10) at different stations will show more dispersion when computed using unfiltered
accelerograms than when using filtered ones. Figure 3b, which shows recordings and
plots of IN(10) for first four events in Table 3, contradicts this. It seems that the
complexity of the sources, at least of large earthquakes, becomes more important when
using filtered traces.
Before we present results of the analysis, we note that the distance between the
near-source station which recorded an event and CU, CU, varies between 142 and 543
km (Table 3). The three events in the distance range of 142 and 220 km are inslab,
normal-faulting earthquakes (H~40 to 60km). Since a detection based on S waves for this
type of earthquake results in less alert time, it may be more useful to develop an
algorithm based on P waves (e.g., Tsuboi et al., 2002; Allen and Kanamori, 2003). In this
study, however, we only consider alerts based on S-wave for even these events. We
reduce computed Arms values of all events to at a common distance of 275 km by
multiplying Arms by (275/ CU). CU for each station is known. Here we are assuming
that the attenuation with distance is reasonably well approximated by (1/CU). As Table 3
shows, the distance between the epicenter and the closest station, S, varies, as does the
75
depth of the focus. In the present analysis, we do not apply any correction to reduce the
computed Arms to a common source-to-station distance. S is greater than 100 km for
events 18, 28, 30, and 35. Although we show these events in the figures, we do not
include them in the statistics.
Figure 4 shows results from the analysis of unfiltered accelerograms with td=10s.
It includes plots of M versus Arms(10), M versus (10), (Amax)CU versus Arms(10), and
(Amax)CU versus (10). The figure includes more than one near-source stations for the
first 7 events in Table 3. As expected, the scatter in M versus Arms(10) plot is large: M
ranges up to two units of magnitude for a given value of Arms(10). We note that if
Arms(10)10 gal, then M5.9 with two exceptions; however, for several events M5.9
even though Arms(10)<10 gal. Thus, if we were to set a public alert threshold at M5.9
and require Arms(10)10 gal to accomplish this, then several events with M5.9 would
go unreported. We would have to set the threshold at Arms(10)1 gal to include all
M5.9 events but, in this case, the alert will also be issued for almost all M<5.9 events as
well. The value of (10) is not a reliable discriminator of M5.9 and M<5.9 events either.
For large earthquakes (10) is expected to be greater than 1. However, for many M5.9
events, (10) is about 1. Quite generally, if (10) 1.35 then M6.9. There are, however,
two earthquakes for which (10) is about 1 although M6.9. One of these events is the
earthquake of 25 April 1989 (event 4, Table 3), which gave rise to (Amax)CU of 17 gal
and was very strongly felt in Mexico City. It follows that a reliable estimation of M is not
possible from Arms(10) and (10). This partly explains the high failure rate of the SAS.
76
Figure 4. Analysis of unfiltered accelerograms with td=10s. Bottom right: M versus Arms(10); Top right:
M versus (10); Bottom left: (Amax)CU versus Arms(10); Top left: (Amax)CU versus (10). Results from
more than one near-source recording are shown for first seven events in Table 3.
Arms(10) value is normalized for a station located at 275 km from CU. Triangle: interplate event;
rectangle: non-interplate event.
77
km). We note that for these events the Amax at CU was greater than 3.5 gal. Although we
do not have the statistics, it is most likely that all of these events were felt by many in the
lake-bed zone and by some in the hill-zone of the city. A general alert would also be
issued for eight other events that gave rise to smaller Amax at CU. These may be
considered false alerts. No alert would have been issued for two events but Amax at CU
during these events was only slightly greater than 3.5 gal. Setting up a lower, second
level of alert is very problematic (Figure 4), unless we require that this level of alert be
issued if 1.0Arms(10)<5.0 gal corresponding to 1(Amax)CU<3.5 gal. This alert,
however, would include all remaining events but one which did not qualify for the
general alert.
We think that only one level of alert, a general alert if Arms(10)5.0 gal, is the
best option. This would include most earthquakes that would be felt by many of the
inhabitants of the city. On average, about one alert/year will be issued which should make
the system credible to the society. Most importantly, no potentially damaging earthquake
would be missed. The fact that some weaker events [(Amax)CU<3.5 gal] will also qualify
for general alert is inevitable. Fortunately, this is an error in the preferable, conservative
direction.
We now consider the consequences of using band-pass filtered (0.2-1.0 Hz) nearsource and CU accelerograms. The results for td =10s are shown in Figure 5. The scatter
in plots of M and (Amax)CU versus Arms(10) plots is somewhat less than the
corresponding plots with unfiltered accelerograms (compare Figures 4 and 5). We now
require Arms (10) 1.0 gal for a general alert. This ensures the alert for events 1 to 7.
This threshold on Arms flags most events with (Amax)CU 2.0 gal. Only three weak
events, (Amax)CU< 2.0 gal, are also flagged for a general alert and one earthquake (event
23, Figure 5) is missed. It is interesting to note that if Arms (10) 1 gal then M 5.9.
78
Setting up a lower, second level of alert presents the same difficulty as in the previous
case. For reasons mentioned earlier, we prefer just one level of alert: a general alert.
We find that the performance of the alert based on filtered traces superior to that
based on unfiltered traces as it decreases the number of false alerts. Table 4 compares the
performance of the proposed general public alert with unfiltered and filtered
accelerograms.
Figure 5. Same as Figure 4 but with band-pass filtered (0.2-1.0 Hz) accelerograms.
Table 4. Performance of proposed public alert for Mexico City for events in Table 3
based on expected (Amax)CU estimated from Arms computed from near-source
accelerograms.
Accelerograms
td
No. of
events to be
flagged
21
Correctly
flagged
events
19
Missed
events
Excess
events
>5.0
Expected
Amax at CU,
gal
>3.5
Unfiltered
10s
Filtered
(0.2-1 Hz)
10s
>1.0
>2.0
18
17
79
The results of the previous section were obtained by processing the raw and
filtered near-source accelerograms using a window of 10s beginning with the arrival of
the S wave. In practice, the detection algorithm will perform continuous analysis of the
signal in a sliding window of 10-s duration. If at any time Arms(10) equals or exceeds a
pre-established threshold (5.0 and 1.0 gal for unfiltered and filtered traces, respectively)
at any sensor, and a nearby sensor confirm the occurrence of an earthquake, then the
system would issue a general public alert. Let us test this algorithm on the dataset of
Table 3.
The results are shown in Figure 6 for unfiltered data. In the figure, the events for
which Arms(10) 5.0 gal at any time over the duration of the accelerogram are shown by
colored symbol (triangle: interplate events; rectangles: non-interplate events). If Arms(10)
remains less than 5.0 gal then the symbol is left blank and no alert is flagged. The colors
are keyed to the time difference between the beginning of the window and the arrival of
the S wave, at the instant the Arms threshold is first reached. In the figure, the Arms(10)
is the maximum value obtained during the processing of the accelerogram. The results are
roughly the same shown in Figure 4: a general alert is issued for all first 19 events in
Table 3, nine false alerts [(Amax)CU<3.5 gal] and two missed alerts [(Amax)CU>3.5 gal].
As seen from the color code, for many events a general alert would have been issued in
less than 10s after S-wave arrival, thus increasing the available warning time for Mexico
City.
The result from processing the filtered data is shown in Figure 7. The Arms
threshold for a general is alert now set at 1.0 gal. All events with (Amax)CU>2.0 gal, but
one, are flagged for a general alert and there three false alerts [(Amax)CU<2 gal]. Again,
in real-time processing, for many events the alert is flagged in less than 10s after the Swave arrival.
80
Figure 6. (Amax)CU versus Arms(10) plot obtained by processing accelerograms of events in Table 3 in the
same manner as the alert algorithm would operate in the field. Continuous analysis is performed on signal
of 10s-long window. When Arms(10)5 gal then a general alert is issued (colored symbol). Expected
(Amax)CU 3.5 gal. The colors are keyed to the beginning of the window with respect to the arrival of the S
wave. If Arms(10) is less than 3.5 gal then both the symbol is left blank.
Figure 7. Same as Figure 6 but with band-pass filtered (0.2-1.0 Hz) accelerograms. In this case, a general
alert is flagged when Arms(10)1 gal. Expected (Amax)CU1.0 gal.
81
As mentioned above and shown in Figure 1, the present SAS array consists of 15
accelerometers located 25 km apart. The array covers a 300 km-long segment of the
Guerrero coast, known as the Guerrero seismic gap. The limited areal coverage of the
SAS excludes the possibility of issuing alerts for earthquakes which occur on the adjacent
segments of the subduction plate boundary and for inslab earthquakes. In fact, the present
coverage of the SAS might not issue an alert if the 1985 Michocan earthquake were to
recur. As Figure 1 shows there are several inslab earthquakes which have given rise to
Amax5 gal at CU since 1964. For such events the sensors of the present array will either
provide no alert or a lower level of alert.
The deployment of an array consisting of three concentric half rings of radii 275,
335, 395 km (Figure 1), with sensors located about 60 km apart in each ring (for a total of
40 sensors for the entire array), could provide an adequate coverage for all coastal and
inland earthquakes which may occur at a distance between 245 and 425 km south of
Mexico City. For such earthquakes, the location of the nearest station will be less than
about 30 km from the epicenter. The depth of the coastal earthquakes ranges between
20km (interplate events) to about 35 km (inslab events). Thus, the hypocentral distance of
the closest sensor for coastal events will less than about 46 km and the S wave will arrive
in less than 13s after the origin. If we allow about 10s for on site processing and
transmission of relevant parameters to the central station in Mexico City, then the alert
can be issued in less than about 23s from the origin. Since the intense S-wave will arrive
in Mexico City after about 80 s from the origin, there will be more than 55s of warning
time for coastal earthquakes. For inslab events, which occur below Mexican altiplano at
depths of 50 to 80 km, the alert time will be somewhat less but still more than 45s.
There still remains the problem of inslab earthquakes which are known to occur
as close as 140 km from Mexico City (Iglesias et al., 2002) and crustal events. Useful
alerts for such events would require additional, closer, half rings of sensors, and a
82
detection algorithm based on P waves (e.g., Tsuboi et al., 2002; Allen and Kanamori,
2003). We have ignored such earthquakes in the present study.
The challenge facing any seismic alert system (SAS) resides in the estimation of
the damaging potential of an earthquake from the analysis of only a few seconds of nearsource ground-motion recordings. The daunting task is made simpler for an alert system
for Mexico City since the city is located more than about 300 km from the locations
where most of the potentially damaging earthquakes occur. This provides reasonable
warning time even when the detection in the near-source region is based on S-wave.
The present SAS for Mexico City has been in operation since August 1991. We
find two basic flaws with the present seismic alert system (SAS) for Mexico City: (1) The
system covers only a part of the region where damaging earthquakes to Mexico City
originate (a well-known limitation of the SAS). (2) The algorithm used by the SAS to
detect potentially damaging earthquakes from the analysis of the near-source groundmotion recordings is inadequate. As a consequence of these two shortcomings, the failure
rate of the SAS is high (Table 1).
In this paper, we have propose a strategy that differs substantially from that
presently implemented by the SAS. It is based on the analysis of near-source recordings
of Mexican earthquakes since 1985 and the corresponding ground motions recorded in
Mexico City. Our approach differs from that of the SAS in one basic aspect: We relate
the near-source ground motion directly to the expected motion in the city, without the
intermediate step of estimating the magnitude of the event. In our proposed scheme, the
alerts would be based on the relationship between root-mean-acceleration (Arms) in the
near-source region and the expected Amax at a reference site in Mexico City, CU. We test
the use of unfiltered and band-pass filtered (0.2-1.0 Hz) accelerograms from near-source
region and Mexico City in the analysis. The choice of the filter is based on the fact that
83
the amplification of seismic waves in the lake-bed zone of the Valley of Mexico occurs in
this frequency band. We find that the use of band-pass filtered near-source accelerograms
of 10-s duration, beginning with the arrival of S-wave, leads to alerts with a much lower
failure rate (Table 4).
The relationship between the level of alert and Amax at CU is a decision which
needs thorough debate by the society. Should public alert be issued only when some
damage is expected in the city? How often should restricted alerts be issued?
We find a single level of alert, a general public alert, as the best option. Our
results suggests that this alert should be issued if Arms at a near-source station, computed
over a 10s-window, exceeds 5.0 gal using the unfiltered signal. For events in Table 3, this
threshold flags all events with Amax at CU greater than 3.5 gal but two, and gives 9 false
alerts (Figure 6). When using the filtered accelerograms, the general public alert should
be issued when Arms exceeds 1 gal. For the events in Table 3, this threshold flags all
events with Amax at CU greater than 2.0 gal but one, and gives only 3 false alerts. For
this reason, we suggest the use of filtered accelerograms in the alert algorithm. In realtime implementation of the algorithm, the alert for many events will be issued in less than
10s after the arrival of S-wave at the station.
To ensure that no important alert to the city is missed and the public credibility in
the system is maintained, it is essential to increase the coverage by the SAS. We suggests
an array consisting of three concentric half rings of radii 275, 335, 395 km (Figure 1),
with sensors located about 60 km apart in each ring (for a total of 40 sensors for the entire
array). Such an array could provide an adequate coverage for all coastal and inland
earthquakes which may occur at a distance between 245 and 425 km south of Mexico
City. If an alert is desired for crustal and inslab earthquakes that may occur nearer to the
city, then an additional closer ring of stations and an algorithm based on P-waves would
be required.
84
Acknowledgements
We express our thanks to Juan Manual Espinosa and other colleagues at CIRES for many
fruitful discussions and for providing us with information. The goal of this research is to
make the SAS, a technical achievement without parallel in Mexico and a fruit of years of
effort by the personnel of CIRES, as useful as possible. The opinion expressed here may
or may not be shared by CIRES. The research was partially funded by CONACyT project
42671-F.
References
Allen. R. M., and H. Kanamori (2003). The potential for earthquake early warning in
southern California, Science 300, 786-789.
Anderson, J.G., J.N. Brune, J. Prince, R. Quaas, S. K. Singh, D. Almora, P. Bodin, M.
Oate, R. Vsquez, and J.M. Velasco (1994). The Guerrero accelerograph network,
Geofsica Intern. 33, 341-372.
Espinosa-Miranda, J.M., A. Uribe, G. Ibarrola, V. Toledo, and C. Rebollar (1989).
Evaluacion de un algoritmo para detectar sismos de subduccin, Mem. VIII Congreso
Nacional de Ingenieria Sismica y VII Congreso Nacional de Ingenieria Estructural,
Acapulco, Mexico, Vol I, A199-A211.
Espinosa-Miranda, J.M., A. Jimnez, G. Ibarrola, F. Alcantar, A. Aguilar, M. Hinostroza,
and S. Maldonado (1995). Mexico City seismic alert system, Seism. Res. Lett. 66, 42-53.
Espinosa-Miranda, J.M. and F. H. Rodrguez (2003). The seismic alert system of Mexico
City, in International Handbook of Earthquake and Engineering Seismology, Vol. 81B,
Academic Press, Chapter 76, 1253-1259.
Iglesias, A., S.K. Singh, J. Pacheco, J.F. Pacheco, and M. Ordaz (2002). A source and
wave propagation study of the Copalillo, Mexico earthquake of July 21, 2000 (Mw=5.9):
Implications for seismic hazard in Mexico City from inslab earthquakes, Bull. Seism. Soc.
Am. 92, 885-895.
Ordaz, M., S.K. Singh, and A. Arciniega (1994). Bayesian attenuation regressions: and
application to Mexico City, Geophys. J. Intern. 117, 335-344.
Singh, S.K., J. Havskov, and L. Astiz (1981). Seismic gaps and recurrence periods of
large earthquakes along the Mexican subduction zone, Bull. Seism. Soc. Am. 71, 827-843.
Singh, S.K., E. Mena, R. Castro, and C. Carmona (1987). Empirical prediction of ground
motion in Mexico City from coastal earthquakes, Bull. Seism. Soc. Am. 77, 1862-1867.
85
86
Conclusiones
87
88
El captulo IV muestra el anlisis del sismo del 18 de Abril del 2002 (Mw=6.8).
Este sismo ocurri muy cerca de la lnea de la trinchera Mesoamericana y present una
deficiencia de energa a altas frecuencias provocando que las aceleraciones, tanto en la
costa como en el resto de las estaciones que lo registraron, fueran especialmente bajas.
Siguiendo un mtodo propuesto anteriormente se clasific al sismo del 18 de Abril como
89
En funcin de los eventos estudiados es razonable pensar que una gran cantidad de
fenmenos ssmicos de diferente naturaleza pudieron haber sido ignorados o simplificados
en el pasado. Este trabajo representa un pequeo esfuerzo para poner en el mismo papel
diferentes aspectos de la sismicidad asociada a la brecha ssmica de Guerrero. Sin embargo,
no se puede dejar de reconocer que la tarea, an pendiente, de compilar todos los
90
trabajos, pasados y futuros, es indispensable para entender de mejor manera el ciclo ssmico
y la sismotectnica de la regin.
Sin poder extrapolar del todo, es viable pensar que en el resto de regiones
ssmicamente activas del pas, la sismicidad est caracterizada por la ocurrencia de
temblores de diversa naturaleza, tal como ocurre en la brecha de Guerrero. La nica manera
de entender mejor los procesos tectnicos que originan dichos temblores as como el riesgo
ssmico asociado a ellos, es a travs de una mayor instrumentacin y esfuerzo cientfico,
tarea, que desde luego, requiere mayores recursos materiales y humanos.
Se debe mencionar que este trabajo fue posible, en gran medida, gracias al apoyo
del Consejo Nacional para la Ciencia y Tecnologa (CONACYT), pero sobretodo, gracias
al esfuerzo de la Universidad Nacional Autnoma de Mxico (UNAM), debido a que en los
momentos en los que el CONACYT decidi dejar de apoyar el proyecto por detalles
burocrticos, la UNAM lo sostuvo.
Sirvan estas lneas para cuando menos remarcar que, al presente ritmo, el futuro no
es prometedor.
91